Plug-ins¶

VVAS is based on the GStreamer framework. This section describes the VVAS GStreamer plug-ins, their input, outputs, and control parameters. The plug-ins source code is available in the vvas-gst-plugins folder of the VVAS source tree. The two types of VVAS GStreamer plug-ins are custom plug-ins and infrastructure plug-ins. Infrastructure plug-ins are developed to enable developers to integrate their kernel into GStreamer based applications without having understanding about GStreamer framework. Infrastructure plug-ins encapsulates most of the basic GStreamer plug-in requirements/features, like memory management, kernel configuration, caps negotiation etc. Same Infrastructure plug-ins can be used for integrating different kernels to realize different functionalities. Custom plug-ins implement a specific functionality, like encode, decode, overlay etc. that can’t be implemented using Infrastructure plug-ins in an optimized way.

This section covers the plug-in that are common for Edge (Embedded) as well as cloud (PCI basede) solutions. There are few plug-ins that are specific to Edge/Embedded platforms and are covered in Plugins for Embedded platforms. Similarly there are few plug-ins that are specific to Cloud/Data Center platforms and these are covered in Plugins for Data Center Platform. The following table lists the VVAS GStreamer plug-ins.

Table 1: GStreamer Plug-ins

Plug-in Name	Functionality
vvas_xmetaaffixer	Plug-in to scale and attach metadata to the frame of different resolutions.
vvas_xabrscaler	Hardware accelerated scaling and color space conversion.
vvas_xmultisrc	A generic infrastructure plug-in: 1 input, N output, supporting transform processing.
vvas_xfilter	A generic infrastructure plug-in: 1 input, 1 output, supporting pass-through, in-place, and transform processing.
vvas_xinfer	An inference plug-in using vvas_xdpuinfer acceleration software library and attaches inference output as GstInferenceMeta to input buffer. Also, this plug-in does optional pre-processing required for inference using vvas_xpreprocess acceleration software library.
vvas_xoptflow	Plug-in to estimate optical flow using previous and current frame.
vvas_xoverlay	Plug-in to draw text, boxes, lines, arrows, circles, polygons, and time stamp on frames.
vvas_xtracker	Plug-in to track objects of interest detected during infer. This plug-in tracks the previously detected objects during the time interval when infer information is not available.
vvas_xmetaconvert	Plugin to convert VVAS ML inference metadata (GstInferenceMeta) into overlay metadata
vvas_xfunnel	Plug-in to serialize data on its sink pads in round robin order
vvas_xdefunnel	Plug-in to de-serialize data serialized using vvas_xfunnel plug-in
vvas_xmulticrop	Hardware accelerated plug-in for doing static and dynamic cropping, pre-processing, scaling, and color space conversion.
vvas_xcompositor	Hardware accelerated N input, 1 output plug-in that combines two or more video frames into a single frame.

Custom Plug-ins¶

There are specific functions, like video decoder, encoder, and meta-affixer where the requirements are difficult to implement in an optimized way using highly simplified and generic infrastructure plug-ins framework. Hence, these functions are implemented using custom GStreamer plug-ins. This section covers details about the custom plug-ins.

vvas_xmetaaffixer¶

The metaaffixer plug-in, vvas_xmetaaffixer, is used to scale the incoming metadata information for the different resolutions. A machine learning (ML) operation can be performed on a different frame resolution and color format than the original frame, but the metadata might be associated with the full resolution, original frame. The vvas_metaaffixer has two types of pads, master pads and slave pads. Input pads are pads on request; the number of input pads can be created based on the number of input sources. There is one mandatory master input pad (sink pad) that receives the original/reference metadata. Other input pads are referred to as slave pads. Metadata received on the master sink pad is scaled in relation to the resolution of each of the slave sink pads. The scaled metadata is attached to the buffer going out of the output (source) slave pads. There can be up to 16 slave pads created as required. For implementation details, refer to vvas_xmetaaffixer source code

Input and Output¶

This plug-in is format agnostic and can accept any input format. It operates only on the metadata. The vvas_xmetaaffixer plug-in supports the GstInferenceMeta data structure. For details about this structure, refer to the VVAS Inference Metadata section.

Control Parameters and Plug-in Properties¶

Table 2: vvas_xmetaaffixer Plug-in Properties

Property Name	Type	Range	Default	Description
sync	Boolean	True/False	True	This property is to enable the synchronization between master and slave pads buffers. If sync=true is set, then the metadata is scaled and attached to buffers on slave pads that have matching PTS or PTS falls within frame duration of the buffer on the master sink pad. If sync=false is set on the element, then the metadata is scaled and attached to all the buffers on the slave pads. If this option is used, there is possibility that the metadata is not suitable for the frames/buffers that are not corresponding to the frames/buffers on the master pad.
timeout	Int64	-1 to 9223372036854	2000	Timeout in millisec. Plug-in will wait for “timeout” duration for a buffer on other pads before pushing buffer on output pad. This is to avoid hang in case buffers are arriving at different frame rates and slave pad buffer is waiting for another buffer at master pad (which may not arrive as there is no more buffers available).

Pad naming syntax¶

The pad naming syntax is listed, and the following image shows the syntax:

MetaAffixer master input pad should be named sink_master.
MetaAffixer master output pad should be named src_master.
MetaAffixer slave input pad should be named sink_slave_0, sink_slave_1.
MetaAffixer slave output pad should be named src_slave_0, src_slave_1, src_slave_2.

Example Pipelines¶

This section covers the example pipelines using the metaaffixer plug-in.

gst-launch-1.0 videotestsrc num-buffers=1 \
! video/x-raw, width=1920, height=1080, format=NV12 \
! queue \
! videoconvert \
! queue \
! ima.sink_master vvas_xmetaaffixer name=ima ima.src_master \
! queue \
! fakesink videotestsrc num-buffers=1 \
! video/x-raw, width=1920, height=1080, format=NV12 \
! queue \
! videoconvert \
! video/x-raw, width=1920, height=1080, format=YUY2 \
! ima.sink_slave_0 ima.src_slave_0 \
! queue \
! fakesink -v

vvas_xabrscaler¶

There are severa use cases where the available frame resolution and color formats may not be suitable for the consumption by the next component. For example, in case of Machine Learning applications, input can be from different sources, resolutions, but ML models work on a fixed resolution. In such cases, the input image needs to be re-sized to a different resolution. Also, the input image color format may be YUV/NV12, but the ML models require image to be in BGR format. In this case we need to do the color space conversion as well. ML model may also require some pre-processing, like Mean Subtraction, Normalization etc. on the input image.

In adaptive bit rate (ABR) use cases, one video is encoded at different bit rates so that it can be streamed in different network bandwidth conditions without any artifacts. To achieve this, input frame is decoded, resized to different resolutions and then re-encoded. vvas_xabrscaler is a plug-in that takes one input frame and can produce several outputs frames having different resolutions and color formats. The vvas_xabrscaler is a GStreamer plug-in developed to accelerate the resize and color space conversion, Mean Subtraction, Normalization, and cropping. For more implementation details, refer to vvas_xabrscaler source code.

This plug-in supports:

Single input, multiple output pads
Color space conversion
Resize
Each output pad has independent resolution and color space conversion capability.
Crop of input buffer

Important

The vvas_xabrscaler plug-in controls the multiscaler kernel. If your application uses this plug-in, then make sure that multi-scaler kernel is included in your hardware design.

Important

Make sure that the multi-scaler hardware kernel supports maximum resolution required by your application.

As a reference, maximum resolution supported by multi-scaler kernel in Smart Model Select example design can be found in multi-scaler kernel config

Prerequisite¶

This plug-in requires the multiscaler kernel to be available in the hardware design and the required color formats are enabled. See Multiscaler Kernel

Input and Output¶

This plug-in accepts buffers with the following color format standards:

RGBx
YUY2
r210
Y410
NV16
NV12
RGB
v308
I422_10LE
GRAY8
NV12_10LE32
BGRx
GRAY10_LE32
BGRx
UYVY
BGR
RGBA
BGRA
I420
GBR

Important

Make sure that the color formats needed for your application are supported by the multi-scaler hardware kernel.

As a reference, multi-scaler configuration for smart model select example design can be found in multi-scaler configuration

Control Parameters and Plug-in Properties¶

The following table lists the GStreamer plug-in properties supported by the vvas_xabrscaler plug-in.

Table 3: vvas_xabrscaler Plug-in Properties

Property Name	Type	Range	Default	Description
avoid-output-copy	Boolean	true/false	False	Avoid output frames copy on all source pads even when downstream does not support GstVideoMeta metadata
enable-pipeline	Boolean	true/false	false	Enable buffer pipelining to improve performance in non zero-copy use cases
in-mem-bank	Unsigned int	0 - 65535	0	VVAS input memory bank to allocate memory
out-mem-bank	Unsigned int	0 - 65535	0	VVAS o/p memory bank to allocate memory
xclbin-location	string	N/A	./binary_container_1 xclbin	The location of xclbin.
kernel-name	string	N/A	v_multi_scaler: multi_scaler_1	Kernel name and instance separated by a colon.
dev-idx	integer	0 to 31	0	Device index This is valid only in PCIe/ Data Center platforms.
ppc	integer	1, 2, 4	2	Pixel per clock supported by a multi- scaler kernel
scale-mode	integer	0, 1, 2	0	Scale algorithm to use: 0:BILINEAR 1:BICUBIC 2:POLYPHASE
coef-load-type	integer	0 => Fixed 1 => Auto	1	Type of filter Coefficients to be used: Fixed or Auto generated
num-taps	integer	6=>6 taps 8=>8 taps 10=>10 taps 12=>12 taps	1	Number of filter taps to be used for scaling
alpha-b	float	0 to 128	0	Mean subtraction for blue channel , needed for PPE
alpha-g	float	0 to 128	0	Mean subtraction for green channel , needed for PPE
alpha-r	float	0 to 128	0	Mean subtraction for red channel , needed for PPE
beta-b	float	0 to 1	1	Scaling for blue channel, needed for PPE
beta-g	float	0 to 1	1	scaling for green channel, needed for PPE
beta-r	float	0 to 1	1	scaling for red channel, needed for PPE
crop-x	unsigned integer	0 to 4294967295	0	Crop X coordinate
crop-y	unsigned integer	0 to 4294967295	0	Crop Y coordinate
crop-width	unsigned integer	0 to 4294967295	0	Crop width ( minimum: 64), if this is 0 or not set, it will be calculated as input width - crop-x
crop-height	unsigned integer	0 to 4294967295	0	Crop height ( minimum: 64), if this is 0 or not set, it will be calculated as input height - crop-y

Note

MultiScaler IP has some alignment requirement, hence user given parameters for crop are aligned as per the IP requirement, alignment ensures that it covers the region of crop specified by user, hence final cropped image may have extra pixels cropped.

Example Pipelines¶

One input one output¶

The following example configures vvas_xabrscaler in one input and one output mode. The input to the scaler is 1280 x 720, NV12 frames that are resized to 640 x 360 resolution, and the color format is changed from NV12 to BGR.

gst-launch-1.0 videotestsrc num-buffers=100 \
! "video/x-raw, width=1280, height=720, format=NV12" \
! vvas_xabrscaler xclbin-location="/usr/lib/dpu.xclbin" kernel-name=v_multi_scaler:{v_multi_scaler_1} \
! "video/x-raw, width=640, height=360, format=BGR" ! fakesink -v

One input multiple output¶

The following example configures vvas_xabrscaler for one input and three outputs. The input is 1920 x 1080 resolution in NV12 format. There are three output formats:

1280 x 720 in BGR format
300 x 300 in RGB format
640 x 480 in NV12 format

gst-launch-1.0 videotestsrc num-buffers=100 \
! "video/x-raw, width=1920, height=1080, format=NV12, framerate=60/1" \
! vvas_xabrscaler xclbin-location="/usr/lib/dpu.xclbin" kernel-name=v_multi_scaler:{v_multi_scaler_1} name=sc sc.src_0 \
! queue \
! "video/x-raw, width=1280, height=720, format=BGR" \
! fakesink sc.src_1 \
! queue \
! "video/x-raw, width=300, height=300, format=RGB" \
! fakesink sc.src_2 \
! queue \
! "video/x-raw, width=640, height=480, format=NV12" \
! fakesink -v

Crop with multiple output:¶

The following example configures vvas_xabrscaler for one input and three outputs. The input is 1920 x 1080 resolution in NV12 format. This input is cropped at X:140, Y:300, Width:640, Height:480.

Cropped input is scaled and converted to below format:

640 * 480 in RGB format
320 * 240 in RGB format
256 * 256 in NV12 format

gst-launch-1.0 -v \
videotestsrc num-buffers=10 ! video/x-raw,format=NV12,width=1920,height=1080 \
! vvas_xabrscaler name=sc dev-idx=0 crop-x=140 crop-y=300 crop-width=640 crop-height=480 \
sc.src_0 ! queue ! video/x-raw,format=RGB,width=640,height=480 ! filesink location=480p.yuv \
sc.src_1 ! queue ! video/x-raw,format=RGB,width=320,height=240 ! filesink location=240p.yuv \
sc.src_2 ! queue ! video/x-raw,format=NV12,width=256,height=256 ! filesink location=256p.yuv -v

vvas_xinfer¶

GStreamer inference plug-in, vvas_xinfer, performs inferencing on video frames/images with the help of Vitis AI library and prepares tree like metadata in GstInferenceMeta object and attaches the same to input GstBuffer. This plug-in triggers optional preprocessing (scale/crop & etc.) operations with the help of vvas_xpreprocessor kernel library (which is on top of Xilinx’s multiscaler IP) on incoming video frames before calling vvas_xdpuinfer kernel library. vvas_xinfer plug-in’s input capabilities are influenced by vvas_xpreprocessor library input capabilities and Vitis AI library capabilities (Vitis-AI does software scaling). If preprocessing is enabled and vvas_xinfer plug-in is receiving non-CMA memory frames, then data copy will be made to ensure CMA frames goes to preprocessing engine. The main advantage of this plug-in is users/customers can realize inference cascading use cases with ease. For implementation details, refer to vvas_xinfer source code

Input and Output¶

Accepts buffers with GRAY8/ NV12/ BGR/ RGB/ YUY2/ r210/ v308/ GRAY10_LE32/ ABGR/ ARGB color formats on input GstPad & output GstPad. Make sure HW supports the required format.
Attaches GstInferenceMeta metadata to output GstBuffer. For details about meta data, refer to VVAS Inference Metadata

Control Parameters¶

Control Parameters¶
Property Name	Type	Range	Default	Description
dynamic-config	String	N/A	Null	String contains dynamic json configuration of inference accelration library
infer-config	String	N/A	Null	location of the inference kernel library configuration file in json format
preprocess-config	String	N/A	Null	location of the scaler kernels config file in json format

infer-config json members¶

Json key	Item	Item description
vvas-library-repo	Description	VVAS libraries repository path to look for kernel libraries by VVAS GStreamer plug-in
	Value type	String
	Mandatory/Optional	Optional
	Default value	/usr/lib in Embedded platforms /opt/xilinx/vvas/lib in PCIe platforms
inference-level	Description	Inference level in cascaded inference use case. e.g. Object detection ML (level-1) followed by object classification (level-2) on detected objects
	Value type	Integer
	Mandatory/Optional	Optional
	Default value	1
low-latency	Description	Parameter to enable/disable low-latency mode in vvas_xinfer and it is useful only when inference-level > 1. If enabled, then vvas_xinfer plug-in will not wait for batch-size frames to be accumulated to reduce latency. If disabled, inference engine can work at maximum throughput.
	Value type	Boolean
	Mandatory/Optional	Optional
	Default value	false
inference-max-queue	Description	Maximum number of input frames those can be queued inside the plug-in. When low-latency is disabled, vvas_xinfer plug-in will wait for inference-max-queue buffers until batch-size is accumulated
	Value type	Integer
	Mandatory/Optional	Optional
	Default value	batch-size
attach-ppe-outbuf	Description	Attaches output of preprocessing library to GstInferenceMeta to avoid redoing of the preprocessing if required.
	Value type	Boolean
	Mandatory/Optional	Optional
	Default value	False
kernel	Description	Kernel object provides information about an VVAS kernel library configuration and kernel library name
	Value type	JSON Object
	Mandatory/Optional	Mandatory
	Default value	None
	Object Members	members of kernel JSON object are mentioned below

infer-config::kernel json members¶

JSON key	Item	Description
library-name	Description	Name of inference kernel library to be loaded for inferencing
	Value type	String
	Mandatory/Optional	Mandatory
	Default value	NULL
config	Description	Inference kernel specific configuration
	Value type	JSON object
	Mandatory/Optional	Mandatory
	Default value	None
	Object members	Contains members specific to inference kernel library. See vvas_xdpuinfer acceleration library for more information

preprocess-config json members¶

Table 4 preprocess-config json members

Json key	Item	Item description
xclbin-location	Description	Location of xclbin which contains scaler IP to program FPGA device based on device-index property
	Value type	String
	Mandatory/Optional	Mandatory
	Default value	NULL
vvas-library-repo	Description	VVAS libraries repository path to look for kernel libraries by VVAS GStreamer plug-in
	Value type	String
	Mandatory/Optional	Optional
	Default value	/usr/lib in Embedded platforms /opt/xilinx/vvas/lib in PCIe platforms
device-index	Description	Device index on which scaler IP is present
	Value type	Integer
	Mandatory/Optional	Mandatory in PCIe platforms In embedded platforms, device-index is not an applicable option as it is always zero
	Default value	-1 in PCIe platforms 0 in Embedded platforms
kernel	Description	Kernel object provides information about an VVAS kernel library configuration and kernel library name
	Value type	JSON Object
	Mandatory/Optional	Mandatory
	Default value	None
	Object Members	members of kernel JSON object are mentioned below

preprocess-config::kernel json members¶

Table 5: preprocess-config::kernel json members

JSON key	Item	Description
library-name	Description	Name of kernel library to be loaded for inferencing
	Value type	String
	Mandatory/Optional	Mandatory
	Default value	NULL
kernel-name	Description	Name of the preprocessing kernel. Syntax : “<kernel_name>:<instance_name>”
	Value type	String
	Mandatory/Optional	Mandatory
	Default value	NULL
config	Description	preprocess kernel specific configuration
	Value type	JSON object
	Mandatory/Optional	Mandatory
	Default value	None
	Object members	Contains members specific to preprocess kernel library

Example infer-json file:

This file is used to configure vvas_xdpuinfer library for Machine Learning.

{
   "vvas-library-repo": "/usr/lib/",
   "inference-level":1,
   "inference-max-queue":30,
   "attach-ppe-outbuf": false,
   "low-latency":false,
   "kernel" : {
       "library-name":"libvvas_xdpuinfer.so",
       "config": {
           "model-name" : "yolov3_voc_tf",
           "model-class" : "YOLOV3",
           "model-format" : "RGB",
           "batch-size":1,
           "model-path" : "/usr/share/vitis_ai_library/models/",
           "run_time_model" : false,
           "need_preprocess" : false,
           "performance_test" : false,
           "debug_level" : 0
       }
   }
}

Example preprocess-config json file:

This json file captures the parameters for the pre-processing kernel.

{
   "xclbin-location":"/usr/lib/dpu.xclbin",
   "vvas-library-repo": "/usr/lib",
   "device-index": 0,
   "kernels" :[
       {
           "kernel-name": "v_multi_scaler:{v_multi_scaler_1}",
           "library-name": "libvvas_xpreprocessor.so",
           "config": {
              "alpha_r" : 0.0,
              "alpha_g" : 0.0,
              "alpha_b" : 0.0,
              "beta_r" : 0.25,
              "beta_g" : 0.25,
              "beta_b" : 0.25,
              "cascade" : 1,
              "debug_level" : 0
           }
       }
   ]

Note

When user wants to perform hardware accelerated pre-processing on input frame using vvas_xinfer plug-in, then the pre-processing parameters provided in preprocess-config json file must be checked for correctness. The pre-processing parameters for the models are provided in the prototxt file for the models. For few models these parameters need to be modified from what it is in the prototxt file. User must provide these modified values in the preprocess-config json file. There are few steps mentioned below to know which models need changes.

To determine the pre-processing parameters for a model, follow the steps mentioned below:

Get the algorithmic scale vector, i.e. “scale” fields for each channel R,G,B mentioned in the model prototxt file. If the model expects input image in RGB format, then the first field will corresponds to Channel R, next will correspond to Channel G, and the last one will be for Channel B. Example prototxt file contents are mentioned below:

model {
        name: "yolov3_voc_416"
        kernel {
            name: "yolov3_voc_416"
            mean: 0.0
            mean: 0.0
            mean: 0.0
            scale: 0.00390625
            scale: 0.00390625
            scale: 0.00390625
}

Calculate the inner scale value from the “fixpos” field of the “input_tensor” by executing the command at command prompt on the target board in case of Embedded platform.

xdputil xmodel -l path-to-xmodel.xmodel

you can get something like:

{
 "subgraphs":[
     {
         "name":"subgraph_data",
         "device":"USER"
     },
     {
        "name":"subgraph_conv1_1",
         "device":"DPU",
         "fingerprint":"0x1000020f6014407",
         "DPU Arch":"DPUCZDX8G_ISA0_B4096_MAX_BG2",
         "input_tensor":[
             {
                 "name":"data_fixed",
                 "shape":"[1, 360, 480, 3]",
                 "fixpos":-1
             }
         ],


 inner scale = 2^fixpos
             = 2^-1
             = 0.5

Multiply the “scale” vector values from prototxt file with the inner scale calculated above. Provide these values in the preprocess-config json file. User need to be careful while assigning the value to the correct channel field.

If the model expects input image in RGB format, then the first scale value in prototxt file will corresponds to channel R. So after multiplying with the "inner scale" value, assign this first value to "beta_r" in preprocess-config json file. Repeat the same for other channels.

Example Simple inference (YOLOv3) pipeline which takes NV12 YUV file (test.nv12) as input is described below:

gst-launch-1.0 filesrc location=<test.nv12> ! videoparse width=1920 height=1080 format=nv12 ! \
vvas_xinfer preprocess-config=yolo_preproc.json infer-config=yolov3_voc.json ! fakesink -v

Example cascade inference (YOLOv3+Resnet18) pipeline which takes NV12 YUV file (test.nv12) as input is described below:

gst-launch-1.0 filesrc location=<test.nv12> ! videoparse width=1920 height=1080 format=nv12 ! \
vvas_xinfer preprocess-config=yolo_preproc.json infer-config=yolov3_voc.json ! queue ! \
vvas_xinfer preprocess-config=resnet_preproc.json infer-config=resnet18.json ! fakesink -v

vvas_xoptflow¶

Optical flow is one of the key functions used in many image processing and computer vision applications like object tracking, motion based segmentation, depth estimation, stitching and video frame rate conversion etc. Optical flow is estimated using previous and current frame pixel information.

For optical flow estimation this plug-in uses hardware accelerator of xfopencv non-pyramid optical flow. This non-pyramid optical flow function takes current and previous frame as input and generates two floating point buffers of x and y direction displacements. Optical flow plug-in attaches these displacement buffers as gstvvasofmeta. For implementation details, refer to vvas_xoptflow source code

../../_images/vvas_optflow_blockdiagram.png

Prerequisite¶

This plug-in uses dense_non_pyr_of_accel kernel. Make sure your xclbin has this kernel.

Input and Output¶

Accepts buffer of NV12 format and generates two metadata buffers of type float and each size equal to the size of frame. For details about the meta data structure, refer to optical_flow_metadata

Plug-in Properties¶

Table 6: vvas_xoptflow Plug-in Properties

Property Name	Type	Range	Default	Description
xclbin-loc	String	NA	./binary_container_1 .xclbin	location of xclbin
dev-idx	Integer	0 to 31	0	device index

example Pipelines¶

The following example demonstrates use of vvas_xoptflow plug-in.

gst-launch-1.0 filesrc location=$1 ! \
h264parse ! \
queue ! \
omxh264dec internal-entropy-buffers=2 ! \
videorate ! video/x-raw, framerate=10/1 ! \
queue ! \
vvas_xinfer preprocess-config=kernel_preprocessor_dev0_yolov3_voc_tf.json infer-config=kernel_yolov3_voc_tf.json name=infer ! \
vvas_xoptflow xclbin-location="/mnt/sd-mmcblk0p1/dpu.xclbin" ! \
fakesink -v

vvas_xoverlay¶

Overlay plug-in is used for drawing bounding boxes, text, lines, arrows, circles, and polygons on frames. Bounding boxes can be drawn by using hardware accelerator. By setting use-bbox-accel property of the plug-in user can use hardware accelerator for drawing bounding boxes. Using display-clock property user can display clock on any part of the frame.

Forr drawing objects like bounding box, lines, arrows etc., the information about these objects must be provided through overlay metadata structure attached to the buffer. For more information about this overlay meta data structure, refer to VVAS Overlay Metadata.

For converting metadata generated from upstream plug-in like infer, opticalflow, segmentation etc., to gstvvasoverlaymeta an intermediate plug-in, vvas_xmetaconvert is to be used before vvas_xoverlay plug-in.

For implementation details, refer to vvas_xoverlay source code

../../_images/vvas_overlay_blockdiagram.png

Supported features¶

Feature	Description
boxes	Drawing bounding boxes with and without background color
text	Overlaying of text with and without background color. Supports text fonts available in opencv
lines	For drawing lines with different thickness and color
arrows	Drawing arrows on either side of line or both the sides
circle	For drawing circles of different radius and thickness without fill
polygons	For drawing closed polygons without fill

Input and Output¶

Supported input buffer formats are RGB, NV12 and Grayscale. Required metadata in gstvvasoverlaymeta type for drawing.

Control Parameters and Plug-in Properties¶

The following table lists the GStreamer plug-in properties supported by the vvas_xoverlay plug-in.

Table 7: vvas_xoverlay Plug-in Properties

Property Name	Type	Range	Default	Description
xclbin-loc	String	NA	./binary_container_1 .xclbin	location of xcllbin
dev-idx	Integer	0 to 31	0	device index
display-clock	Boolean	0 or 1	0	flag for indicating displaying clock
use-bbox-accel	Boolean	0 or 1	0	flag to use accelerator for bounding boxes
clock-fontname	Integer	0 or 7	0	font number for clock based on opencv
clock-fontscale	float	0.1 to 1.0	0.5	font scale of display clock
clock-fontcolor	Integer	0 to 4,294,967,295	0xff00	RGB color of display clock as 0xRGB
clock-xoffset	Integer	0 to frame width	100	x starting position
clock-yoffset	Integer	0 to frame height	50	y starting position

Example Pipelines¶

The following example demonstrates use of vvas_xoverlay plug-in with vvas_xxinfer and vvas_xmetaconvert plug-ins for drawing bounding boxes. vvas_xinfer plug-in produces inference result and stores in VVAS_GstInference data structure. This metadata is parsed and translated into a different metadata structure that understood by vvas_xmetaconvert plug-in to draw the bounding box. This translation of meta data is done by vvas_xmetaconvert plug-in.

gst-launch-1.0 filesrc location="<PATH>/001.bgr" blocksize=150528 numbuffers=1
! videoparse width=224 height=224 framerate=30/1 format=16
! vvas_xfilter name="kernel1" kernels-config="<PATH>/kernel_resnet50.json"
! vvas_xmetaconvert config-location="<PATH>/metaconvert.json"
! vvas_xoverlay ! filesink location=output.bgr

vvas_xtracker¶

The tracker plug-in tracks detected objects in a sequence of video frames. Because tracking is easier than detection, tracking algorithms can use fewer computational resources than running an object detector on every frame and hence finds many applications.

Each tracker element in the frame is given a unique ID and is initialized with the coordinates of the object detected by ML block. The object is snipped from the frame and taken as a template. Afterward, the tracker element updates the filter every time that there is an object hit, adapting itself to the possible changes which an object might have during the video.

The tracker plug-in supports three types of object tracking algorithms, namely Intersection-Over-Union (IOU), Kernelized Correlation Filters (KCF) and Minimum Output Sum of Squared Error (MOSSE). IOU algorithm is used when detection output is available for every frame. Kernelized Correlation Filters (KCF) algorithm or Minimum Output Sum of Squared Error (MOSSE) filter algorithm implemented completely in software using optimizations for the host processor. For implementation details, refer to vvas_xtracker source code

Tracker plug-in supports detecting 16 objects supporting tracker Id from 0 to 15. Beyond these, the tracker Ids restarts from 0.

../../_images/vvas_tracker_blockdiagram.png

In the above pipeline, preprocessor generates two outputs, one in the resolution and color format required by the object detection model and the other one in the resolution required by the model in NV12 format. One output is provided to ML block to detect the object in the frame. Since the tracker is used in the pipeline, ML block can be run at lower framerate using videorate plug-in. The detection co-ordinates of the frames detected by the ML block are used by tracker plug-in to initialize the objects to track by tracker plug-in. For the frames that are skipped by ML block, tracker plug-in tracks and updates the tracked co-ordinates in the metadata for the next plug-in to draw the bounding boxes.

Input and Output¶

Accepts buffers with NV12 color formats on input GstPad & output GstPad.

Control Parameters and Plug-in Properties¶

The following table lists the GStreamer plug-in properties supported by the vvas_xtracker plug-in.

Table 8: vvas_xtracker Plug-in Properties

Property Name	Type	Range	Default	Description
tracker-algo	Enum	0, 1, 2	0	Tracker algorithm to use 0 - Tracker IOU algorithm 1 - Tracker MOSSE algorithm 2 - Tracker KCF algorithm
IOU-with-color	Boolean	0 or 1	0	To use color information during matching 0 - No color information for matching 1 - Use color information for matching
obj-match-color-space	Enum	0 or 1	1	Color space to use for objects matching 0 - RGB color space for object match 1 - HSV color space for object match
feature-length	Enum	22, 31	31	Object feature length (required only for KCF algorithm)
search-scale	Enum	0, 1, 2	0	Scales to verify for object localization 0 - Search all scales (up, down and same scale) 1 - Search up and same scale 2 - Search down and same scale
detection-interval	Integer	0 to 4294967295	5	Object detection interval in number of frames i.e frame interval at which objects are passed via ML block for detection. If video framerate is 60fps and ML block framerate is 12 fps then set the detection-interval to 60/12 = 5 Please make sure this field is set correctly to get accurate tracker resutls
inactive-wait- interval	Integer	0 to 4294967295	24	Number of frames to wait before stopping tracking of inactive objects
min-object-width	Integer	0 to 4294967295	20	Minimum object width in pixels to consider for tracking
min-object-height	Integer	0 to 4294967295	60	Minimum object height in pixels to consider for tracking
max-object-width	Integer	0 to 4294967295	200	Objects width above maximum width are not considered for tracking
max-object-height	Integer	0 to 4294967295	360	Objects height above maximum height are not considered for tracking
num-frames-confidence	Integer	0 to 4294967295	3	Number of times object need to be detected continuously to consider for tracking
match-search-region	Float	1.0 to 2.0	1.0	Object search region during detection for IOU based matching
relative-search- region	Float	1.0 to 2.5	1.5	Object search region with respect to detection coordinates
correlation-threshold	Float	0.1 to 1.0	0.7	Object correlation threshold value for matching
overlap-threshold	Float	0.0011 to 1.0	0.001	Object overlap threshold to consider for matching
scale-change- threshold	Float	0.1 to 1.0	0.7	Maximum object scale change threshold to consider for matching. Value of 1 means double the scale
correlation-weight	Float	0.0 to 1.0	0.6	Weightage for correlation value
overlap-weight	Float	0.0 to 1.0	0.3	Weightage for overlap value
scale-change-weight	Float	0.0 to 1.0	0.1	Weightage for change in scale
occlusion-threshold	Float	0.0 to 1.0	0.4	Threshold for considering object as occluded
confidence-score- threshold	Float	0.0 to 1.0	0.25	Confidence score of tracker to be consider for tracking objects

Example Pipelines¶

The following example demonstrates use of vvas_xtracker plug-in along with detection from ML at interval of 5.

gst-launch-1.0 filesrc location=$1 ! \
h264parse ! queue ! \
omxh264dec internal-entropy-buffers=3 ! \
queue ! \
tee name=t t. ! \
queue ! \
vvas_xabrscaler xclbin-location="/media/sd-mmcblk0p1/dpu.xclbin" kernel-name=v_multi_scaler:{v_multi_scaler_1} name=sc sc.src_0 ! \
queue ! \
videorate ! \
video/x-raw, framerate=6/1 ! \
queue ! \
vvas_xinfer preprocess-config=kernel_preprocessor_refinedet_pruned_0_96.json infer-config=kernel_refinedet_pruned_0_96.json name=infer ! \
queue ! \
mftrack1.sink_master vvas_xmetaaffixer name=mftrack1 mftrack1.src_master ! \
queue  ! \
fakesink sc.src_1 ! \
queue ! \
mftrack1.sink_slave_0 mftrack1.src_slave_0 ! \
queue ! \
vvas_xtracker tracker-algo=2 ! \
queue max-size-bytes=0 min-threshold-buffers=1 max-size-buffers=2 max-size-time=0 ! \
mftrack2.sink_master vvas_xmetaaffixer name=mftrack2 mftrack2.src_master ! \
queue  ! \
fakesink t. ! \
queue ! \
mftrack2.sink_slave_0 mftrack2.src_slave_0 ! \
queue ! \
vvas_xmetaconvert config-location=metaconvert.json ! \
queue ! \
vvas_xoverlay ! \
queue ! \
kmssink driver-name="xlnx" fullscreen-overlay=1 sync=false -v

vvas_xmetaconvert¶

GStreamer vvas_xmetaconvert plug-in converts VVAS ML inference metadata (GstInferenceMeta) into overlay metadata structure needed by vvas_overlay plug-in. Configuration parameters which are frequently used are added as plug-in properties as well (Ex: “font”, “font-size”, “thickness” etc). For implementation details, refer to vvas_xmetaconvert source code

JSON File for vvas_xmetaconvert¶

This section describes the JSON file format and configuration parameters for the vvas_xmetaconvert plug-in.

The following example is of a JSON file to pass to the vvas_xmetaconvert.

{
  "config": {
    "font_size" : 0.5,
    "font" : 3,
    "thickness" : 2,
    "mask_level" : 0,
    "label_color" : { "blue" : 0, "green" : 0, "red" : 0 },
    "y_offset" : 0,
    "label_filter" : [ "class", "probability" ],
    "classes" : [
      {
        "name" : "car",
        "blue" : 255,
        "green" : 0,
        "red"  : 0,
        "masking"  : 0
      },
      {
        "name" : "person",
        "blue" : 0,
        "green" : 255,
        "red"  : 0,
        "masking"  : 0
      },
      {
        "name" : "bus",
        "blue" : 0,
        "green" : 0,
        "red"  : 255,
        "masking"  : 0
      },
      {
        "name" : "bicycle",
        "blue" : 0,
        "green" : 0,
        "red"  : 255,
        "masking"  : 0
      }
    ]
  }
}

Various configuration parameters used by vvas_xmetaconvert are described in the following table.

Table 9: vvas_xmetaconvert Parameters

Parameter	Expected Values	Description
font	0 to 7	Font for the label text. 0: Hershey Simplex 1: Hershey Plain 2: Hershey Duplex 3: Hershey Complex 4: Hershey Triplex 5: Hershey Complex Small 6: Hershey Script Simplex 7: Hershey Script Complex
font_size	0.5 to 1	Font fraction scale factor that is multiplied by the font-specific base size.
thickness	Integer 1 to 3	The thickness of the line that makes up the rectangle. Negative values, like -1, signify that the function draws a filled rectangle. The recommended value is between 1 and 3.
mask_level	Integer	In case of cascaded ML pipeline, user can use this field to mask out the results of a particular level. All bounding boxes in that level will be masked with black color. When set to 0, none of the levels are masked
label_color	{ “blue” : 0, “green” : 0, “red” : 0 }	The color of the label is specified.
label_filter	[ “class”, “probability” ]	This field indicates that all information printed is the label string. Using “class” alone adds the ML classification name. For example, car, person, etc. The addition of “probability” in the array adds the probability of a positive object identification.
y_offset	Integer 0 to height	‘y’ offset to be added along height for label in case of classification model
classes	{ “name” : “car”, “blue” : 255, “green” : 0, “red” : 0, “masking” : 0 }	This is a filtering option when using the vvas_xboundingbox. The bounding box is only drawn for the classes that are listed in this configuration. Other classes are ignored. For instance, if “car”, “person”, “bicycle” is entered under “classes”, then the bounding box is only drawn for these three classes, and other classes like horse, motorbike, etc. are ignored. The expected value columns show an example of how each class should be described. All objects in this example, by class, are drawn using the color combination listed. The class names in this list matches the class names assigned by the vvas_xdpuinfer. Otherwise, the bounding box is not drawn. “masking” flag can be used to mask all objects of a class Set it to 1 for enabling masking. This flag will override “mask_level” field. For instance, if “mask_level” is set 1 and “masking” for “car” is set 1 and if level 1 has cars and busses, then only cars are masked not the entire level. For face detect, keep the “classes” array empty.
display_level	Integer 0 to N 0 => all levels N => specific level	display bounding box of one particular level or all levels

Example Pipelines¶

The following example demonstrates use of vvas_xmetaconvert with vvas_overlay plug-in for drawing bounding boxes.

gst-launch-1.0 filesrc location="<PATH>/001.bgr" blocksize=150528 numbuffers=1
! videoparse width=224 height=224 framerate=30/1 format=16
! vvas_xfilter name="kernel1" kernels-config="<PATH>/kernel_resnet50.json"
! vvas_xmetaconvert config-location="<PATH>/metaconvert.json"
! vvas_overlay ! filesink location=output.bgr

vvas_xfunnel¶

vvas_xfunnel gstreamer plug-in serializes data on its sink pads, it iterates all the sink pads in round robin order, if sink pad has buffer available; it sends it on the output pad, else the plug-in will wait for preset (user configurable) time, by that time also if data is not available the plug-in will skip that sink pad and it will probe next sink pad.

Whenever new sink pad is added to or removed from this plug-in, it sends custom events to notify vvas_xdefunnel to create or destroy source pads. Metadata on each buffer is attached to enable vvas_xdefunnel plug-in to decide the source pad to which that buffer has to be sent.

vvas_xfunnel supports all caps.

Refer vvas_xdefunnel plug-in which de-serializes these serialized data and pushes them to the output pads. For vvas_xfunnel implementation details, refer to vvas_xfunnel source code

This plug-in supports:

Multiple input, single output pad
Sink pads are request pads and they can be added or removed dynamically.
Serialization of parallel data using round robin algorithm

Important

Caps of all the sink pads must be same, this is to avoid caps renegotiation always when sending buffer from the sink pad which has different caps.

Important

There are several custom events sent from this plugin for vvas_xdefunnel plug-in, intermediate plug-ins must pass these events.

Control Parameters and Plug-in Properties¶

The following table lists the GStreamer plug-in properties supported by the vvas_xfunnel plug-in.

Table 10: vvas_xfunnel Plug-in Properties

Property Name	Type	Range	Default	Description
queue-size	unsigned integer	1 - 100	2	queue size for each sink pad
sink-wait-timeout	unsigned integer	1 - 100	33	Time to wait before switching to next sink pad in mili seconds, default will be calculated as 1/FPS

Example Pipelines¶

The pipeline mentioned below is using vvass_xvcudec and vvas_xvcuenc plug-ins that is supported on PCIe/Data center platform. To execute the similar pipeline on Embedded platform, use the decoder and encoder plug-ins supported on Embedded platform. For details, refer to OMX Encoder/Decoder Plug-in

gst-launch-1.0 \
filesrc location= video1.h264 ! h264parse ! vvas_xvcudec  dev-idx=0 ! fun.sink_0 \
filesrc location= video2.h264 ! h264parse ! vvas_xvcudec  dev-idx=0 ! fun.sink_1 \
filesrc location= video3.h264 ! h264parse ! vvas_xvcudec  dev-idx=0 ! fun.sink_2 \
vvas_xfunnel name=fun ! identity silent=false ! vvas_xdefunnel name=def \
def.src_0 ! queue ! vvas_xvcuenc  dev-idx=0 ! h264parse ! filesink async=false sync=false location= out1.h264 \
def.src_1 ! queue ! vvas_xvcuenc  dev-idx=0 ! h264parse ! filesink async=false sync=false location= out1.h264 \
def.src_2 ! queue ! vvas_xvcuenc  dev-idx=0 ! h264parse ! filesink async=false sync=false location= out1.h264 -v

vvas_xdefunnel¶

vvas_xdefunnel gstreamer plug-in de-serializes data coming from vvas_xfunnel plug-in. This plug-in creates and destroys source pads based on the events from the vvas_xfunnel plug-in. For vvas_xdefunnel implementation details, refer to vvas_xdefunnel source code

This plug-in supports:

Single input, multiple output pad
Source pads are sometimes pad.
Source pads can be added and removed dynamically.

vvas_xdefunnel supports all caps.

Control Parameters and Plug-in Signals¶

The following table lists the GStreamer plug-in signals supported by the vvas_xdefunnel plug-in.

Table 11: vvas_xdefunnel Plug-in signals

Signal Name	Description	callback
pad-added	This signal will be emitted when new source pad is created.	void user_function (GstElement* object, GstPad* arg0, gpointer user_data);
pad-removed	This signal will be emitted when a source pad is removed.	void user_function (GstElement* object, GstPad* arg0, gpointer user_data);

Example Pipelines¶

The pipeline mentioned below is using vvass_xvcudec and vvas_xvcuenc plug-ins that is supported on PCIe/Data center platform. To execute the similar pipeline on Embedded platform, use the decoder and encoder plug-ins supported on Embedded platform. For details, refer to OMX Encoder/Decoder Plug-in

gst-launch-1.0 \
filesrc location= video1.h264 ! h264parse ! vvas_xvcudec  dev-idx=0 ! fun.sink_0 \
filesrc location= video2.h264 ! h264parse ! vvas_xvcudec  dev-idx=0 ! fun.sink_1 \
filesrc location= video3.h264 ! h264parse ! vvas_xvcudec  dev-idx=0 ! fun.sink_2 \
vvas_xfunnel name=fun ! identity silent=false ! vvas_xdefunnel name=def \
def.src_0 ! queue ! vvas_xvcuenc  dev-idx=0 ! h264parse ! filesink async=false sync=false location= out1.h264 \
def.src_1 ! queue ! vvas_xvcuenc  dev-idx=0 ! h264parse ! filesink async=false sync=false location= out1.h264 \
def.src_2 ! queue ! vvas_xvcuenc  dev-idx=0 ! h264parse ! filesink async=false sync=false location= out1.h264 -v

vvas_xmulticrop¶

In ML applications we often need to crop the video frame for processing objects and number of objects per frame may differ. For ML we may also need preprocessing.

vvas_xmulticrop is a Gstreamer plug-in to do scaling, color space conversion, preprocessing and to crop single or multiple objects. It takes one input frame and can crop one or more objects from it. Scaling, color space conversion and preprocessing can be applied to all the cropped objects/buffers. If user wants to get cropped buffer directly on the output pad, he can use static cropping, but with static cropping user can crop only one object. If user wants to crop more than one objects, s/he must go with dynamic cropping, dynamically cropped objects/buffers will not be sent onto the output pad, they will be attached as metadata into the output buffer. Both static and dynamic cropping can be performed simultaneously.

For more implementation details, refer to vvas_xmulticrop source code.

This plug-in supports:

Single input, single output pad
Color space conversion
Resize
Pre-processing
Static and Dynamic cropping

vvas_xmulticrop plug-in is similar to vvas_xabrscaler excepts below features:

It has only one output pad
It supports dynamic cropping.

Static Cropping: For cropping only one object. The cropped object/buffer will be sent to the output pad. s-crop-x, s-crop-y, s-crop-width, s-crop-height are the properties to set the crop parameters.

Dynamic Cropping: For cropping more than one objects from the input buffer. To use dynamic crop feature user must send crop coordinates attached to the buffer in GstVideoRegionOfInterestMeta metadata and GstVideoRegionOfInterestMeta->roi_type must be set to “roi-crop-meta”. One instance of GstVideoRegionOfInterestMeta in buffer represents one object for dynamic crop. vvas_xmulticrop will dynamically crop the object and attach the cropped objects/buffers to output buffer in GstVideoRegionOfInterestMeta->params. GstVideoRegionOfInterestMeta->params is a GList * of GstStructure. Dynamically cropped buffer is attached into this field, The name of GstStructure is “roi-buffer”, and this GstStructure has only one field “sub-buffer” of type GST_TYPE_BUFFER. User should extract this cropped buffer use it and unref it. User can choose to resize these cropped buffers to one width and height by setting d-height and d-width properties. If these properties are not set, then cropped buffers will not be resized.

If user wants cropped buffers to be of different format than the input format, he can specify this by setting d-format gstreamer property. If this property is not set all cropped buffers will have format same as input buffer.

User specified pre-processing will be applied to dynamically cropped buffers only, if user wants it to be applied on output buffer/static crop buffers also, he can set ppe-on-main buffer gstreamer property.

Static and dynamic cropping both is possible simultaneously. As explained statically cropped buffer will be sent on the output pad and dynamically cropped buffers will be attached into that output buffer.

vvas_xmulticrop supports at max 39 dynamic crops. Memory for dynamically cropped buffers is allocated from a Gstreamer buffer pool, there is no upper limit on this buffer pool. So, if buffers are not freed, new buffers will be allocated which may lead to more memory consumption.

Important

The vvas_xmulticrop plug-in controls the multiscaler kernel. If your application uses this plug-in, then make sure that multi-scaler kernel is included in your hardware design.

Important

Make sure that the multi-scaler hardware kernel supports maximum resolution required by your application.

As a reference, maximum resolution supported by multi-scaler kernel in Smart Model Select example design can be found in multi-scaler kernel config

Prerequisite¶

This plug-in requires the multiscaler kernel to be available in the hardware design. See Multiscaler Kernel

Input and Output¶

This plug-in accepts buffers with the following color format standards:

RGBx
YUY2
r210
Y410
NV16
NV12
RGB
v308
I422_10LE
GRAY8
NV12_10LE32
BGRx
GRAY10_LE32
BGRx
UYVY
BGR
RGBA
BGRA
I420
GBR

Important

Make sure that the color formats needed for your application are supported by the multi-scaler hardware kernel.

As a reference, multi-scaler configuration for smart model select example design can be found in multi-scaler configuration

Control Parameters and Plug-in Properties¶

The following table lists the GStreamer plug-in properties supported by the vvas_xmulticrop plug-in.

Table 12: vvas_xmulticrop Plug-in Properties

Property Name	Type	Range	Default	Description
avoid-output-copy	Boolean	true/false	False	Avoid output frames copy even when downstream does not support GstVideoMeta metadata
enable-pipeline	Boolean	true/false	false	Enable buffer pipelining to improve performance in non zero-copy use cases
in-mem-bank	Unsigned int	0 - 65535	0	VVAS input memory bank to allocate memory
out-mem-bank	Unsigned int	0 - 65535	0	VVAS o/p memory bank to allocate memory
xclbin-location	string	N/A	./binary_container_1 xclbin	The location of xclbin.
kernel-name	string	N/A	v_multi_scaler: multi_scaler_1	Kernel name and instance separated by a colon.
dev-idx	integer	0 to 31	0	Device index This is valid only in PCIe/ Data Center platforms.
ppc	integer	1, 2, 4	2	Pixel per clock supported by a multi- scaler kernel
scale-mode	integer	0, 1, 2	0	Scale algorithm to use: 0:BILINEAR 1:BICUBIC 2:POLYPHASE
coef-load-type	integer	0 => Fixed 1 => Auto	1	Type of filter Coefficients to be used: Fixed or Auto generated
num-taps	integer	6=>6 taps 8=>8 taps 10=>10 taps 12=>12 taps	1	Number of filter taps to be used for scaling
alpha-b	float	0 to 128	0	Mean subtraction for blue channel , needed only if PPE is required
alpha-g	float	0 to 128	0	Mean subtraction for green channel , needed only if PPE is required
alpha-r	float	0 to 128	0	Mean subtraction for red channel , needed only if PPE is required
beta-b	float	0 to 1	1	Scaling for blue channel, needed only if PPE is required
beta-g	float	0 to 1	1	scaling for green channel, needed only if PPE is required
beta-r	float	0 to 1	1	scaling for red channel, needed only if PPE is required
s-crop-x	unsigned int	0 to 4294967925	0	Crop X coordinate for static cropping
s-crop-y	unsigned int	0 to 4294967925	0	Crop Y coordinate for static cropping
s-crop-width	unsigned int	0 to 4294967925	0	Crop width for static cropping (minimum: 64), when this is 0 or not set, it will be calculated as input width - s-crop-x
s-crop-height	unsigned int	0 to 4294967925	0	Crop height for static cropping (minimum: 64), when this is 0 or not set, it will be calculated as input height - s-crop-y
d-width	unsigned int	0 to 4294967925	0	Width of dynamically cropped buffers
d-height	unsigned int	0 to 4294967925	0	Height of dynamically cropped buffers
d-format	integer	0,2,4,5,7,8, 11,12,15,16, 23,25,41,45, 48,51,78,79, 83	0	Format of dynamically cropped buffers
ppe-on-main-buffer	boolean	true/false	0	Apply pre- processing on main buffer also

Note

MultiScaler IP has some alignment requirement, hence user given parameters for crop are aligned as per the IP requirement, alignment ensures that it covers the region of crop specified by user, hence final cropped image may have extra pixels cropped. Crop width and height must be at least 64.

Example Pipelines¶

The pipeline mentioned below is for PCIe/Data Center platform. In case you want to execute this pipeline on Embedded platform, then remove dev-idx property in the pipelines mentioned below.

Below pipeline converts NV12 to RGB and performs scaling from 1920x1080 to 640x480. The pipeline mentioned below is for PCIe/Data Center platform.

gst-launch-1.0 -v \
videotestsrc num-buffers=10 ! video/x-raw,format=NV12,width=1920,height=1080 \
! vvas_xmulticrop dev-idx=0 \
! video/x-raw,format=RGB,width=640,height=480 ! filesink location=out.yuv

Below pipeline performs pre-processing along with color space conversion and scaling on output buffers

gst-launch-1.0 -v \
videotestsrc num-buffers=10 ! video/x-raw,format=NV12,width=1920,height=1080 \
! vvas_xmulticrop dev-idx=0 ppe-on-main-buffer=true alpha-r=124 alpha-g=116 alpha-b=104 beta-r=0.547 beta-g=0.56 beta-b=0.557 \
! video/x-raw,format=RGB,width=640,height=480 ! filesink location=out.yuv

Below pipeline performs static cropping at (x,y) = (100,80) and (width,height)= (1280,720), this cropped buffers gets scaled to 640x480 and converted to RGB.

gst-launch-1.0 -v \
videotestsrc num-buffers=10 ! video/x-raw,format=NV12,width=1920,height=1080 \
! vvas_xmulticrop dev-idx=0 s-crop-x=100 s-crop-y=80 s-crop-width=1280 s-crop-height=720 \
! video/x-raw,format=RGB,width=640,height=480 ! filesink location=out.yuv

Below code shows how to add GstVideoRegionOfInterestMeta for dynamic cropping.

GstVideoRegionOfInterestMeta *meta;
meta = gst_buffer_add_video_region_of_interest_meta (buffer, "roi-crop-meta", 0, 0, 0, 0);
if (meta) {
  meta->id = id;
  meta->parent_id = p_id;
  meta->x =  x;
  meta->y =  y;
  meta->w = w;
  meta->h = h;
  printf("meta: x:%u y:%u, w:%u h:%u", meta->x, meta->y, meta->w, meta->h);
}

Below code shows how to read GstVideoRegionOfInterestMeta and how to extract dynamically cropped buffer/object

read_crop_meta (GstBuffer *buf) {
  gpointer state = NULL;
  GstMeta *_meta;
  while ((_meta = gst_buffer_iterate_meta_filtered (buf, &state,
              GST_VIDEO_REGION_OF_INTEREST_META_API_TYPE))) {
    GstStructure *s = NULL;
    GstVideoRegionOfInterestMeta *roi_meta =
                          (GstVideoRegionOfInterestMeta  *) _meta;
    if (g_strcmp0("roi-crop-meta", g_quark_to_string (roi_meta->roi_type))) {
      //This is not the metadata we are looking for
      continue;
    }
    //Got the ROI crop metadata, prepare output buffer
    //Extract dynamically cropped buffer from this meta
    s = gst_video_region_of_interest_meta_get_param (roi_meta, "roi-buffer");
    if (s) {
      GstBuffer *sub_buffer = NULL;
      gst_structure_get (s, "sub-buffer", GST_TYPE_BUFFER, &sub_buffer, NULL);
      if (sub_buffer) {
        //use sub_buffer and unref it
        dump_dynamically_cropped_buffer (sub_buffer);
      } else {
        printf("couldn't get sub buffer");
      }
    } else {
      printf("couldn't get expected struct");
    }
  }
  return TRUE;
}

Below code shows how to read/dump dynamically cropped buffers

dump_dynamically_cropped_buffer (GstBuffer *sub_buffer) {
  /* Read GstVideoMeta from the buffer, dump the buffer to file */
  GstVideoMeta *vmeta = NULL;
  FILE *fp;
  GstMapInfo map = {0};
  gchar name[256] = {0};
  GstBuffer *new_outbuf;
  GstVideoFrame new_frame = { 0 }, out_frame = { 0 };
  GstVideoInfo *vinfo;

  if (!sub_buffer) {
    return FALSE;
  }

  vmeta = gst_buffer_get_video_meta (sub_buffer);
  if (!vmeta) {
    printf"couldn't get video meta");
  }

  sprintf (name, "dynbuf_%ux%u_%s.yuv", vmeta->width, vmeta->height,
                        gst_video_format_to_string (vmeta->format));
  fp = fopen (name, "wb");
  if (!fp) {
    return FALSE;
  }

  vinfo = gst_video_info_new ();
  gst_video_info_set_format (vinfo, vmeta->format, vmeta->width, vmeta->height);

  new_outbuf = gst_buffer_new_and_alloc (GST_VIDEO_INFO_SIZE (vinfo));
  if (!new_outbuf) {
    printf("couldn't allocate buffer");
    gst_video_info_free (vinfo);
    fclose (fp);
    return FALSE;
  }

  gst_video_frame_map (&out_frame, vinfo, sub_buffer, GST_MAP_READ);
  gst_video_frame_map (&new_frame, vinfo, new_outbuf, GST_MAP_WRITE);
  gst_video_frame_copy (&new_frame, &out_frame);
  gst_video_frame_unmap (&out_frame);
  gst_video_frame_unmap (&new_frame);
  gst_video_info_free (vinfo);

  if (gst_buffer_map ( new_outbuf, &map, GST_MAP_READ)) {
    fwrite (map.data, map.size, 1, fp);
  }
  gst_buffer_unmap (new_outbuf, &map);
  gst_buffer_unref (new_outbuf);
  fclose (fp);
  gst_buffer_unref (sub_buffer);
  return TRUE;
}

vvas_xcompositor¶

vvas_xcompositor is a hardware accelerated N input, 1 output Gstreamer plugin that combines two or more video frames into a single frame. It can accept the below mentioned video formats. For each of the requested sink pads it will compare the incoming geometry and framerate to define the output parameters. Indeed, output video frames will have the geometry of the biggest incoming video stream and the framerate of the fastest incoming one.

In case input and output formats are different, then the color space conversion will be hardware accelerated by vvas_xcompositor. For implementation details, refer to vvas_xcompositor source code

Input and Output¶

This plug-in accepts buffers with the following color format standards:

RGBx
YUY2
r210
Y410
NV16
NV12
RGB
v308
I422_10LE
GRAY8
NV12_10LE32
BGRx
GRAY10_LE32
BGRx
UYVY
BGR
RGBA
BGRA
I420
GBR

Important

Make sure that the color formats needed for your application are supported by the multi-scaler hardware kernel.

As a reference, multi-scaler configuration for smart model select example design can be found in multi-scaler configuration

Individual parameters for each input stream can be configured on the GstCompositorPad:¶

“xpos”: The x-co-ordinate position of the top-left corner of the current frame in output buffer.
“ypos”: The y-co-ordinate position of the top-left corner of the current frame in output buffer.
“width”: The width of the current picture in the output buffer; If the input width and the width in output buffer are different, then hardware accelerated resize operation will be performed.
“height”: The height of the current picture in the output buffer; If the input height and the height in output buffer are different, then hardware accelerated resize operation will be performed.
“zorder”: The z-order position of the picture in the composition.

Plugin properties¶

Table 13: vvas_xcompositor Plug-in Properties

Property Name	Type	Range	Default	Description
avoid-output-copy	Boolean	true or false	false	Avoid output frames copy on all source pads even when downstream does not support GstVideoMeta metadata
best-fit	Boolean	true or false	false	downscale/upscale the input video to best-fit in each window
dev-idx	Integer	-1 to 31	-1	Device index Valid only for PCIe/ ta Center platforms
enable-pipeline	Boolean	true or false	false	Enable buffer pipelining to improve performance in non zero-copy use cases
in-mem-bank	Unsigned Integer	0 - 65535	0	VVAS input memory bank to allocate memory
out-mem-bank	Unsigned Integer	0 - 65535	0	VVAS output memory bank to allocate memory
ppc	Integer	1, 2, 4	2	Pixel per clock supported by a multi-scaler kernel
scale-mode	Integer	0, 1, 2	0	Scale algorithm to use: 0:BILINEAR 1:BICUBIC 2:POLYPHASE
coef-load-type	Integer	0 => Fixed 1 => Auto	1	Type of filter Coefficients to be used: Fixed or Auto generated
num-taps	Integer	6=>6 taps 8=>8 taps 10=>10 taps 12=>12 taps	1	Number of filter taps to be used for scaling
kernel-name	String	NA	v_multi_scaler :{v_multi_scaler_1}	String defining the kernel name and instance as mentioned in xclbin
xclbin-location	String	NA	NULL	Location of the xclbin to program devices

vvas_xcompositor pad properties¶

Table 14: vvas_xcompositor Pad Properties

Property Name	Type	Range	Default	Description
xpos	Unsigned Integer	0 to 2147483647	0	The x-co-ordinate position of the top-left corner of the current farme in output buffer.
ypos	Unsigned Integer	0 to 2147483647	0	The y-co-ordinate position of the top-left corner of the current farme in output buffer.
height	Integer	-1 to 2147483647	-1	The height of the current picture in the output buffer; If the input height and the height in output buffer are different, then hardware accelerated resize operation will be performed. Setting default/-1 treats o/p height as input height
width	Integer	-1 to 2147483647	-1	The width of the current picture in the output buffer; If the input width and the width in output buffer are different, then hardware accelerated resize operation will be performed. Setting default/-1 treats o/p width as input width
zorder	Unsigned Integer	-1 to 16	-1	The z-order positon of the picture in the composition.

The example pipeline with vvas_xcompositor plug-in is as mentioned below.

The pipeline mentioned below is using omxh264dec plug-in and this plug-in is supported on embedded platforms only. For PCIe/Data center platform, use vvas_xvcudec plug-in.

#! /bin/bash

PAD_PROPERTIES="\
sink_0::xpos=0 sink_0::ypos=0 \
sink_1::xpos=1920 sink_1::ypos=0 \
sink_2::xpos=0 sink_2::ypos=1080 \
sink_3::xpos=1920 sink_3::ypos=1080 \
"

gst-launch-1.0 -v  filesrc location=$1 !\
h264parse !\
omxh264dec !\
comp.sink_0 \
vvas_xcompositor xclbin-location=/usr/lib/dpu.xclbin $PAD_PROPERTIES name=comp !\
video/x-raw , width=3840, height=2160 , format=NV12  !\
queue !\
fpsdisplaysink video-sink="kmssink  bus-id=a0130000.v_mix async=true" text-overlay=false sync=false \
filesrc location=$2 !\
h264parse !\
omxh264dec name=decoder_1 !\
queue !\
comp.sink_1 \
filesrc location=$3 !\
h264parse !\
omxh264dec !\
queue !\
comp.sink_2 \
filesrc location=$4 !\
h264parse !\
omxh264dec !\
queue !\
comp.sink_3

Infrastructure Plug-ins and Acceleration Software Libraries¶

Infrastructure plug-ins are generic plug-ins that interact with the acceleration kernel through a set of APIs exposed by an acceleration software library corresponding to that kernel. Infrastructure plug-ins abstract the core/common functionality of the GStreamer framework (for example: caps negotiation and buffer management).

Table 15: Infrastructure Plug-ins

Infrastructure Plug-ins	Function
vvas_xfilter	Plug-in has one input, one output Support Transform, passthrough and inplace transform operations
vvas_xmultisrc	Plug-in support one input and multiple output pads. Support transform operation

Note

Though one input and one output kernel can be integrated using any of the two infrastructure plug-ins, we recommend using vvas_xfilter plugin for one input and one output kernels.

Acceleration software libraries control the acceleration kernel, like register programming, or any other core logic required to implement the functions. Acceleration software libraries expose a simplified interface that is called by the GStreamer infrastructure plug-ins to interact with the acceleration kernel. The following table lists the acceleration software libraries developed to implement specific functionality. These libraries are used with one of the infrastructure plug-ins to use the functionality a GStreamer-based application. Example pipelines with GStreamer infrastructure plug-ins and acceleration software libraries are covered later in this section.

Table 16: Acceleration Software Libraries

Acceleration Software Library	Function
vvas_xdpuinfer	Library based on Vitis AI to control DPU kernels for machine learning.
vvas_xboundingbox	Library to draw a bounding box and labels on the frame using OpenCV.

GStreamer infrastructure plug-ins are available in the vvas-gst-plugins folder in the vvas source code tree. The following section describes each infrastructure plug-in.

vvas_xfilter¶

The GStreamer vvas_xfilter is an infrastructure plug-in that is derived from GstBaseTransform. It supports one input pad and one output pad. The vvas_xfilter efficiently works with hard-kernel/soft-kernel/software (user-space) acceleration software library types as shown in the following figure.

This plug-in can operate in the following three modes.

Passthrough: Useful when the acceleration software library does not need to alter the input buffer.
In-place: Useful when the acceleration software library needs to alter the input buffer.
Transform: In this mode, for each input buffer, a new output buffer is produced.

You must set the mode using the JSON file. Refer to JSON File Schema for information related to the kernels-config property.

The vvas_xfilter plug-in takes configuration file as one of the input properties, kernels- config. This configuration file is in JSON format and contains information required by the kernel. During initialization, the vvas_xfilter parses the JSON file and performs the following tasks:

Finds the VVAS acceleration software library in the path and loads the shared library.
Understands the acceleration software library type and prepares the acceleration software library handle (VVASKernel) to be passed to the core APIs.

Input and Output¶

The vvas_xfilter accepts the buffers with the following color formats on input GstPad and output GstPad.

GRAY8
NV12
BGR
RGB
YUY2
r210
v308
GRAY10_LE32
ABGR
ARGB

The formats listed are the Xilinx IP supported color formats. To add other color formats, update the vvas_kernel.h and vvas_xfilter plug-ins.

Control Parameters and Plug-in Properties¶

The following table lists the GObject properties exposed by the vvas_xfilter. Most of them are only available in PCIe based platforms.

Table 17: GObject Properties

Property Name	Platforms Supported	Type	Range	Default	Description
dynamic-config	PCIe and Embedded	String	N/A	Null	JSON formatted string contains kernel specific configuration for run time changes
dev-idx	PCIe only	Integer	0 to 31	0	Device used for kernel processing, xclbin download.
kernels-config	PCIe and Embedded	String	N/A	NULL	JSON configuration file path based on VVAS acceleration software library requirements. Refer to the JSON File Schema
sk-cur-idx	PCIe only	Integer	0 to 31	0	Softkernel current index to be used for executing job on device.
reservation-id	PCIe only	Integer	0 to 1024	0	Reservation ID provided by the Xilinx resource manager (XRM).

JSON Format for vvas_xfilter Plug-in¶

The following table provides the JSON keys accepted by the GStreamer vvas_xfilter plug-in.

Table 18: Root JSON Object Members

JSON Key	Item	Item Description
xclbin-location	Description	The location of the xclbin that is used to program the FPGA device.
	Value type	String
	Mandatory or optional	Conditionally mandatory: If the VVAS acceleration software library is developed for hard-kernel IP and soft-kernel, then the xclbin-location is mandatory. If the VVAS acceleration software library is developed for a software library (e.g., OpenCV), then the xclbin-location is not required
	Default value	NULL
vvas-library-repo	Description	This is the VVAS libraries repository path to look for acceleration software libraries by the VVAS GStreamer plug-in.
	Value type	String
	Mandatory or optional	Optional
	Default value	/usr/lib
element-mode	Description	GStreamer element mode to operate. Based on your requirement, choose one of the following modes: Passthrough: In this mode, element does not want to alter the input buffer. Inplace: In this mode, element wants to alter the input buffer itself instead of producing new output buffers. Transform: In this mode, element produces a output buffer for each input buffer.
	Value type	Enum
	Mandatory or optional	Mandatory
kernels	Description	This is the array of kernel objects. Each kernel object provides information about an VVAS acceleration software library configuration. The vvas_xfilter only takes the first kernel object in the kernel array.
	Value type	Array of objects
	Mandatory or optional	Mandatory
	Default value	None
	Minimum value	1
	Maximum value	10
	Object members	Refer to Kernel JSON Object

The information in the following table is specific to the kernel chosen.

Table 19: Kernel JSON Object Members

JSON Key	Item	Item Description
library-name	Description	The name of the VVAS acceleration software library to be loaded by the VVAS GStreamer plug-ins. The absolute path of the kernel library is formed by the pre-pending vvas-library-repo path.
	Value type	String
	Mandatory or optional	Mandatory
	Default value	None
kernel-name	Description	The name of the IP/kernel in the form of <kernel name>:<instance name>
	Value type	String
	Mandatory or optional	Optional
	Default value	None
kernel-access-mode	Description	Mode in which IP/Kernel is to be accessed. “shared” or “exclusive”
	Value type	String
	Mandatory or optional	Optional
	Default value	“shared”
config	Description	Holds the configuration specific to the VVAS acceleration software library. VVAS GStreamer plug-ins do not parse this JSON object, instead it is sent to the acceleration software library.
	Value type	Object
	Mandatory or optional	Optional
	Default value	None
softkernel	Description	Contains the information specific to the soft-kernel. This JSON object is valid only for the PCIe based platforms.
	Value type	Object
	Mandatory or optional	Mandatory if kernel library is written for soft-kernel.
	Default value	None
	Members	Not required for embedded platforms.

For vvas_xfilter implementation details, refer to vvas_xfilter source code

Example JSON Configuration Files¶

JSON File for vvas_xdpuinfer library¶

The following JSON file is for pure software-based acceleration, it does not involve any kernel for acceleration. However, the Vitis AI API based DPU is a special case, where the DPU hardware kernel is controlled by Vitis AI. The acceleration software library calls the Vitis AI API calls and it is implemented as a pure software acceleration software library. There is no need to provide the path of the xclbin in the JSON file.

{
   "vvas-library-repo": "/usr/lib/",
   "element-mode":"inplace",
   "kernels" :[
      {
         "library-name":"libvvas_xdpuinfer.so",
         "config": {
            "model-name" : "densebox_320_320",
            "model-class" : "FACEDETECT",
            "model-format": ""BGR",
            "model-path" : "/usr/share/vitis_ai_library/models/",
            "run_time_model" : false,
            "need_preprocess" : true,
            "performance_test" : true,
            "max_num" : -1,
            "prob_cutoff" : 0.0,
            "debug_level" : 1
         }
      }
   ]
}

JSON File for a Hard Kernel¶

The following JSON file uses vvas_xfilter to control multi-scaler IP (hard-kernel). The acceleration software library accessing the register map is libvvas_xcrop.so.

{
   "xclbin-location":"/usr/lib/dpu.xclbin",
   "vvas-library-repo": "/usr/lib/",
   "element-mode":"passthrough",
   "kernels" :[
       {
          "kernel-name":"v_multi_scaler:{v_multi_scaler_1}",
          "library-name":"libvvas_xcrop.so",
          "config": {
          }
       }
   ]
}

Example Pipelines¶

This section covers the GStreamer example pipelines using the vvas_xfilter infrastructure plug-in and several acceleration software libraries. This section covers the bounding box functionality and the machine learning functions using the vvas_xfilter plug-in.

The bounding box functionality is implemented in the vvas_xboundingbox acceleration software library that is controlled by the vvas_xfilter plug-in.
Machine learning using the DPU is implemented by the vvas_xdpuinfer acceleration software library that is called by the vvas_xfilter plug-in.

Bounding Box Example¶

This section describes how to draw a bounding box and label information using the VVAS infrastructure plug-in vvas_xfilter and the vvas_xboundingbox acceleration software library. The vvas_xboundingbox interprets machine learning inference results from the vvas_xdpuinfer acceleration software library and uses an OpenCV library to draw the bounding box and label on the identified objects.

For vvas_xboundingbox implementation details, refer to vvas_xboundingbox source code

../../_images/X24998-vvas-xboundingbox.png

Prerequisites¶

There are a few requirements before start running bounding box examples. Make sure these prerequisites are met.

Installation of OpenCV: The vvas_xboundingbox uses OpenCV for the graphics back-end library to draw the boxes and labels. Install the libopencv-core-dev package (the preferred version is 3.2.0 or later).

JSON File for vvas_xboundingbox¶

This section describes the JSON file format and configuration parameters for the bounding box acceleration software library. The GStreamer vvas_xfilter plug-in used in the in-place mode. Bounding box and labels are drawn on identified objects on the incoming frame. Bounding box functionality is implemented in the libvvas_xboundingbox.so acceleration software library.

The following example is of a JSON file to pass to the vvas_xfilter.

{
   "xclbin-location":"/usr/lib/dpu.xclbin",
   "vvas-library-repo": "/usr/local/lib/vvas",
   "element-mode":"inplace",
   "kernels" :[
      {
         "library-name":"libvvas_xboundingbox.so",
         "config": {
            "font_size" : 0.5,
            "font" : 3,
            "thickness" : 2,
            "debug_level" : 2,
            "label_color" : { "blue" : 0, "green" : 0, "red" : 0 },
            "label_filter" : [ "class", "probability" ],
            "classes" : [
               {
                  "name" : "car",
                  "blue" : 255,
                  "green" : 0,
                  "red" : 0
               },
               {
                  "name" : "person",
                  "blue" : 0,
                  "green" : 255,
                  "red" : 0
               },
               {
                  "name" : "bicycle",
                  "blue" : 0,
                  "green" : 0,
                  "red" : 255
               }
            ]
         }
      }
   ]
}

Various configuration parameters of the bounding box acceleration software library are described in the following table.

Table 20: vvas_xboundingbox Parameters

Parameter	Expected Values	Description
debug_level	0: LOG_LEVEL_ERROR 1: LOG_LEVEL_WARNING 2: LOG_LEVEL_INFO 3: LOG_LEVEL_DEBUG	Enables the log levels. There are four log levels listed in the expected values column.
font	0 to 7	Font for the label text. 0: Hershey Simplex 1: Hershey Plain 2: Hershey Duplex 3: Hershey Complex 4: Hershey Triplex 5: Hershey Complex Small 6: Hershey Script Simplex 7: Hershey Script Complex
font_size	0.5 to 1	Font fraction scale factor that is multiplied by the font-specific base size.
thickness	Integer 1 to 3	The thickness of the line that makes up the rectangle. Negative values, like -1, signify that the function draws a filled rectangle. The recommended value is between 1 and 3.
label_color	{ “blue” : 0, “green” : 0, “red” : 0 }	The color of the label is specified.
label_filter	[ “class”, “probability” ]	This field indicates that all information printed is the label string. Using “class” alone adds the ML classification name. For example, car, person, etc. The addition of “probability” in the array adds the probability of a positive object identification.
classes	{ “name” : “car”, “blue” : 255, “green” : 0, “red” : 0 }	This is a filtering option when using the vvas_xboundingbox. The bounding box is only drawn for the classes that are listed in this configuration. Other classes are ignored. For instance, if “car”, “person”, “bicycle” is entered under “classes”, then the bounding box is only drawn for these three classes, and other classes like horse, motorbike, etc. are ignored. The expected value columns show an example of how each class should be described. All objects in this example, by class, are drawn using the color combination listed. The class names in this list matches the class names assigned by the vvas_xdpuinfer. Otherwise, the bounding box is not drawn. For face detect, keep the “classes” array empty.
display_level	Integer 0 to N 0 => all levels N => specific level	display bounding box of one particular level or all levels

An example of using a bounding box along with the machine learning plug-in is shown in the Multi Channel ML Tutorial.

Machine Learning Example¶

This section discusses how machine learning solutions can be implemented using the VVAS infrastructure vvas_xfilter plug-in and the vvas_xdpuinfer acceleration software library. Please follow vvas_xinfer to run vvas_xdpuinfer with inference plug-in.

The vvas_xdpuinfer is the acceleration software library that controls the DPU through the Vitis AI interface. The vvas_xdpuinfer does not modify the contents of the input buffer. The input buffer is passed to the Vitis AI model library that generates the inference data. This inference data is then mapped into the VVAS meta data structure and attached to the input buffer. The same input buffer is then pushed to the downstream plug-in.

For vvas_xdpuinfer implementation details, refer to vvas_xdpuinfer source code

Prerequisites¶

There are a few requirements to be met before you can start running the machine learning examples.

Model Information¶

The model directory name must match with the ELF and prototxt files inside the model directory. The model directory should contain:

model-name.elf/model-name.xmodel and model-name.prototxt file.
label.json file containing the label information, if the models generate labels.

The following is an example of the model directory (yolov2_voc), which contains the yolov2_voc.xmodel and yolov2_voc.prototxt files along with the label.json file.

xclbin Location¶

By default, the Vitis AI interface expects xclbin to be located at /usr/lib/ and the xclbin is called dpu.xclbin. Another option is to use the environment variable XLNX_VART_FIRMWARE to specify the path of xclbin using export XLNX_VART_FIRMWARE=/where/your/dpu.xclbin.

Input Image¶

The vvas_xdpuinfer works with raw BGR and RGB images as required by the model. Make sure you have specified correct color format in model-format field in json file. The exact resolution of the image to vvas_xdpuinfer must be provided, it is expected by the model. There is a performance loss if a different resolution of the BGR image is provided to the vvas_xdpuinfer because resizing is done on the CPU inside the Vitis AI library.

JSON File for vvas_xdpuinfer¶

The following table shows the JSON file format and configuration parameters for vvas_xdpuinfer.

Table 21: JSON File for vvas_xdpuinfer

Parameter	Type	Expected Values	Default	Description
xclbin-location	string	This field is not needed for dpuinfer	NULL	By default, Vitis AI expects xclbin to be located at /usr/lib/ and xclbin is called “dpu.xclbin”. The environment variable XLNX_VART_FIRMWARE could also be used to change the path. For example, export XLNX_VART_FIRMWARE=/where/your/dpu.xclbin. Alternatively, you may also provide the xclbin path in `/etc/vart.conf`. In this release, `/etc/vart.conf` has “/media/sd-mmcblk0p1/dpu.xclbin”
vvas-library-repo	string	`/usr/local/lib/vvas/`	`usr/lib/`	This is the path where the vvas_xfilter will search the acceleration software library. The kernel name is specified in the library-name parameter of the JSON file.
element-mode	string	inplace	None	Because the input buffer is not modified by the ML operation, but the metadata generated out of an inference operation needs to be added/appended to the input buffer, the GstBuffer is writable. The vvas_xfilter is configured in inplace mode
kernels	N/A	N/A	N/A	The JSON tag for starting the kernel specific configurations.
kernel-name	string	N/A	NULL	The name and instance of a kernel separated by “:”
library-name	string	N/A	NULL	Acceleration software library name for the kernel. It is appended to the vvas-l ibrary-repo for an absolute path.
config	N/A	N/A	N/A	The JSON tag for kernel-specific configurations depending on the acceleration software library.
model-name	string	resnet50	N/A	Name string of the machine learning model to be executed. The name string should be same as the name of the directory available in model -path parameter file. If the name of the model ELF file is resnet50.elf, then the model-name is resnet50 in the JSON file. The ELF file present in the specified path model-path of the JSON file.
model-class	string	YOLOV3 FACEDETECT CLASSIFICATION SSD REFINEDET TFSSD YOLOV2	N/A	Class of some model corresponding to model. Some examples are shown below: YOLOV3: yolov3_adas_pruned_0_9, yolov3_voc, yolov3_voc_tf FACEDETECT: densebox_320_320, densebox_640_360 CLASSIFICATION: resnet18, resnet50, resnet_v1_50_tf
model-format	string	RGB/BGR	N/A	Image color format required by model.
model-path	string	`/usr/share/vitis_ai_library/models/`	N/A	Path of the folder where the model to be executed is stored.
need_preprocess	Boolean	True/False	True	If need_preprocess = true: Normalize with mean/scale through the Vitis AI Library If need_preprocess = false: Normalize with mean/scale is performed before feeding the frame to vvas_xdpuinfer. The Vitis AI library does not perform these operations.
performance_test	Boolean	True/False	False	Enable performance test and corresponding flops per second (f/s) display logs. Calculates and displays the f/s of the standalone DPU after every second.
debug_level	integer	0 to 3	1	Used to enable log levels. There are four log levels for a message sent by the kernel library code, starting from level 0 and decreasing in severity till level 3 the lowest log-level identifier. When a log level is set, it acts as a filter, where only messages with a log-level lower than it, (therefore messages with an higher severity) are displayed. 0: This is the highest level in order of severity: it is used for messages about critical errors, both hardware and software related. 1: This level is used in situations where you attention is immediately required. 2: This is the log level used for information messages about the action performed by the kernel and output of model. 3: This level is used for debugging.

Model Parameters¶

The Vitis AI library provides a way to read model parameters by reading the configuration file. It facilitates uniform configuration management of model parameters. The configuration file is located in the model-path of the JSON file with [model_name].prototxt. These parameters are model specific. For more information on model parameters, refer to the Vitis AI Library User Guide (UG1354).

Example GStreamer Pipelines¶

This section describes a few example GStreamer pipelines.

Classification Example Using Resnet50¶

The following pipeline performs ML using a Resnet50 model. DPU configuration uses kernels- config = ./json_files/kernel_resnet50.json for the vvas_xdpuinfer. The output of the vvas_xfilter is passed to fakesink along with the metadata.

../../_images/example-using-resnet50-model.png

The GStreamer command for the example pipeline:

gst-launch-1.0 filesrc location="<PATH>/001.bgr" blocksize=150528 numbuffers=1
! videoparse width=224 height=224 framerate=30/1 format=16
! vvas_xfilter name="kernel1" kernels-config="<PATH>/kernel_resnet50.json"
! fakesink

The JSON file for the vvas_xdpuinfer to execute resnet50 model-based classification pipeline is described below.

{
   "vvas-library-repo": "/usr/local/lib/vvas/",
   "element-mode":"inplace",
   "kernels" :[
      {
         "library-name":"libvvas_xdpuinfer.so",
         "config": {
            "model-name" : "resnet50",
            "model-class" : "CLASSIFICATION",
            "model-format : "BGR"
            "model-path" : "/usr/share/vitis_ai_library/models/",
            "run_time_model" : false,
            "need_preprocess" : true,
            "performance_test" : true,
            "debug_level" : 2
         }
      }
   ]
}

Note

If “need_preprocess” = false, then pre-processing operations like, Normalization, scaling must be performed on the frame before feeding to vvas_xfilter/vvas_xdpuinfer otherwise results may not be as expected.

vvas_xmultisrc¶

The GStreamer vvas_xmultisrc plug-in can have one input pad and multiple-output pads. This single plug-in supports multiple acceleration kernels, each controlled by a separate acceleration software library.

For vvas_xmultisrc implementation details, refer to vvas_xmultisrc source code

Input and Output¶

Input and output accept buffers with the following color formats on input GstPad and output GstPad.

GRAY8
NV12
BGR
RGB
YUY2
r210
v308
GRAY10_LE32
ABGR
ARGB

The formats listed are the Xilinx IP supported color formats. To add other color formats, update the vvas_kernel.h and vvas_xfilter plug-ins.

Control Parameters and Plug-in Properties¶

Table 22: Plug-in Properties

Property Name	Type	Range	Default	Description
kconfig	String	N/A	NULL	Path of the JSON configuration file based on the VVAS acceleration software library requirements. For further information, refer to JSON File Schema
dynamic-config	String	N/A	NULL

JSON format for vvas_xmultisrc¶

This section covers the format of JSON file/string to be passed to the vvas_xmultisrc plug-in.

Example JSON File¶

The following example file describes how to specify two kernels that are being controlled by the single instance of the vvas_xmultisrc plug-in. Modify this json file as per your kernels and acceleration software library. The next section describes each field in this example file.

{
   "xclbin-location":"/usr/lib/binary_1.xclbin",
   "vvas-library-repo": "/usr/lib/",
   "kernels" :[
      {
         "kernel-name":"resize:resize_1", <------------------ kernel 1
         "library-name":"libvvas_xresize.so",
         "config": {
            x : 4,
            y : 7
         }
      }
      {
         "kernel-name":"cvt_rgb:cvt_rgb_1", <-------------- kernel 2
         "library-name":"libcvt_bgr.so",
         "config": {
            name = "xilinx",
            value = 98.34
         }
      }
   ]
}

Table 23: JSON Properties

Property Name	Type	Range	Default	Description
xclbin-location	String	N/A	NULL	The path of xclbin including the xclbin name. The plug-in downloads this xclbin and creates an XRT handle for memory allocation and programming kernels.
vvas-library-repo	String	N/A	/usr/lib	The library path for the VVAS repository for all the acceleration software libraries.
kernels	N/A	N/A	N/A	The JSON tag for starting the kernel specific configurations.
kernel-name	String	N/A	NULL	Name and instance of a kernel separated by “:” as mentioned in xclbin.
library-name	String	N/A	NULL	The acceleration software library name for the kernel. This is appended to vvas-library-repo for an absolute path.
kernel-access-mode	String	N/A	“shared”	Mode in which IP/Kernel is to be accessed. “shared or “exclusive”
config	N/A	N/A	N/A	The JSON tag for kernel specific configurations that depends on the acceleration software library.

Source Pad Naming Syntax¶

For single output pad naming is optional. For multiple pads, the source pads names shall be as mentioned below, assuming the name of the plug-in as sc.

sc.src_0, sc.src_1 …..sc.src_n

Example Pipelines¶

Single Output Pad¶

The following example demonstrates the vvas_xmultisrc plug-in configured for one input and one output. A test video pattern is generated by the videotestrc plug-in and passed to the vvas_xmultisrc plug-in. Depending on the kernel being used, vvas_xmultisrc uses kernel.json to configure the kernel for processing the input frame before passing it to the fakesink.

gst-launch-1.0 videotestsrc
! "video/x-raw, width=1280, height=720, format=BGR"
! vvas_xmultisrc kconfig="/root/jsons/<kernel.json>"
! "video/x-raw, width=640, height=360, format=BGR"
! fakesink -v

The following is an example kernel.json file having mean_value and use_mean as kernel configuration parameters. Modify this as per your kernel requirements.

{
   "xclbin-location": "/usr/lib/dpu.xclbin",
   "vvas-library-repo": "/usr/lib/vvas",
   "kernels": [
      {
         "kernel-name": "<kernel-name>",
         "library-name": "libvvas_xresize_bgr.so",
         "config": {
         "use_mean": 1,
         "mean_value": 128
         }
      }
   ]
}

GstVvasBufferPool¶

The GStreamer VVAS buffer pool holds the pool of video frames allocated using the GStreamer allocator object. It is derived from the GStreamer base video buffer pool object (GstVideoBufferPool).

The VVAS buffer pool:

Allocates buffers with stride and height alignment requirements. (e.g., the video codec unit (VCU) requires the stride to be aligned with 256 bytes and the height aligned with 64 bytes)
Provides a callback to the GStreamer plug-in when the buffer comes back to the pool after it is used.

The following API is used to create the buffer pool.

GstBufferPool *gst_vvas_buffer_pool_new (guint stride_align, guint
elevation_align)

Parameters:
   stride_align - stride alignment required
   elevation_align - height alignment required

Return:
   GstBufferPool object handle

Plug-ins can use the following API to set the callback function on the VVAS buffer pool and the callback function is called when the buffer arrives back to the pool after it is used.

Buffer release callback function pointer declaration:
typedef void (*ReleaseBufferCallback)(GstBuffer *buf, gpointer user_data);

void gst_vvas_buffer_pool_set_release_buffer_cb (GstVvasBufferPool *xpool,
ReleaseBufferCallback release_buf_cb, gpointer user_data)

Parameters:
   xpool - VVAS buffer pool created using gst_vvas_buffer_pool_new
   release_buf_cb - function pointer of callback
   user_data - user provided handle to be sent as function argument while
calling release_buf_cb()

Return:
    None

GstVvasAllocator¶

The GStreamer VVAS allocator object is derived from an open source GstAllocator object designed for allocating memory using XRT APIs. The VVAS allocator is the backbone to the VVAS framework achieving zero-copy (wherever possible).

Allocator APIs¶

GStreamer plug-in developers can use the following APIs to interact with the VVAS allocator. To allocate memory using XRT, the GStreamer plug-ins and buffer pools require the GstAllocator object provided by the following API.

GstAllocator* gst_vvas_allocator_new (guint dev_idx, gboolean need_dma)

Parameters:
   dev_idx - FPGA Device index on which memory is going to allocated
   need_dma - will decide memory allocated is dmabuf or not

Return:
   GstAllocator handle on success or NULL on failure

..note:: In PCIe platforms, the DMA buffer allocation support is not available. This means that the need_dma argument to gst_vvas_allocator_new() API must be false.

Use the following API to check if a particular GstMemory object is allocated using GstVvasAlloctor.

gboolean gst_is_vvas_memory (GstMemory *mem)

Parameters:
   mem - memory to be validated

Return:
   true if memory is allocated using VVAS Allocator or false

When GStreamer plug-ins are interacting with hard-kernel/IP or soft-kernel, the plug-ins need physical memory addresses on an FPGA using the following API.

guint64 gst_vvas_allocator_get_paddr (GstMemory *mem)

Parameters:
   mem - memory to get physical address

Return:
   valid physical address on FPGA device or 0 on failure

Use the following API when plug-ins need an XRT buffer object (BO) corresponding to an VVAS memory object.

guint gst_vvas_allocator_get_bo (GstMemory *mem)

Parameters:
   mem - memory to get XRT BO

Return:
   valid XRT BO on success or 0 on failure

JSON Schema¶

This section covers the JSON schema for the configuration files used by the infrastructure plug-ins. For more details, refer to JSON Schema

VVAS Inference Meta Data¶

This section covers details about inference meta data structure used to store meta data. For more details, refer to VVAS Inference Metadata

Overlay Meta Data¶

vvas_xoverlay plug-in understand GstOverlayMetaData structure. Details about this structure can be referred at VVAS Overlay Metadata.