Tool/software:
Hello Team,
We have followed below command to run the edge AI-Dataflow on DSP processor on our Custom device based on AM62A7 processor.
We are using "AR0235" camera sensor module with our custom board. For that reason we have modified the "Object Detection" command minor based on camera-sensor compatibility.
Modified-Command for object detection using DSP model:-
media-ctl -V '"ar0235 2-0036":0 [fmt:SRGGB8_1X8/1920x1200 field:none]' gst-launch-1.0 -v v4l2src device=/dev/video-rpi-cam0 io-mode=5 ! queue leaky=2 ! video/x-bayer, width=1920, height=1200, format=rggb ! tiovxisp dcc-isp-file=/opt/imaging/ar0235/dcc_viss_8bit_1920x1200.bin format-msb=7 sink_0::dcc-2a-file=/opt/imaging/ar0235/dcc_2a_8bit_1920x1200.bin sink_0::device=/dev/v4l-rpi-subdev0 ! video/x-raw, format=NV12 ! tiovxmultiscaler name=split_01 src_0::roi-startx=0 src_0::roi-starty=0 src_0::roi-width=1920 src_0::roi-height=1200 target=0 split_01. ! queue ! video/x-raw, width=480, height=416 ! tiovxmultiscaler target=1 ! video/x-raw, width=416, height=416 ! tiovxdlpreproc model=/media/userdata/model_zoo/ONR-OD-8200-yolox-nano-lite-mmdet-coco-416x416 out-pool-size=4 ! application/x-tensor-tiovx ! tidlinferer target=1 model=/media/userdata/model_zoo/ONR-OD-8200-yolox-nano-lite-mmdet-coco-416x416 ! post_0.tensor split_01. ! queue ! video/x-raw, width=1280, height=720 ! post_0.sink tidlpostproc name=post_0 model=/media/userdata/model_zoo/ONR-OD-8200-yolox-nano-lite-mmdet-coco-416x416 alpha=0.4 viz-threshold=0.6 top-N=5 display-model=true ! queue ! mosaic_0. tiovxmosaic name=mosaic_0 target=1 src::pool-size=4 sink_0::startx="<320>" sink_0::starty="<150>" sink_0::widths="<1280>" sink_0::heights="<720>" ! video/x-raw,format=NV12, width=1920, height=1080 ! v4l2h264enc extra-controls="controls, frame_level_rate_control_enable=1, video_bitrate=10000000, video_gop_size=30" ! h264parse ! matroskamux ! tcpserversink host=172.25.3.171 port=5000
We are observing random failure while running the above command on our custom board. Please check the screenshot below for the same. Also observed that the command works(success) randomly on board, but that success ratio is very low, and most of the time we are observing failure and board freeze issues when running the above command on board.
Issue:- ssh might get disconnected or memory freezing issue can be seen on terminal. In such cases, Device needs to be rebooted. We need to hard reboot device multiple if above command fails to execute.
Important Note:- . Rootfs used for custom-board is minimal due to the limited size of eMMC on our custom board. We have added a minimal additional package to run the available AI model from TI on DSP.
Please check above details and let us know how to resolve above issue and make this command stable on custom board.
We need to run above command on our custom-board with stability to check the DSP load using "perf_stats" utility.
Please provide fix on above issue asap.
Thanks,
Nisarg
