How to use onnx file with deepstream-test1-usbcam + Custom models

deepstream_test_1_usb_2.py (15.2 KB)

this test app just run with one model, try with deepstream-test2

I know but I modified it just like test2
Primary detector and tracker works.

try add: process-mode=2 to sgie config

Adding process-mode=2 causes Segmentation fault on first detection

can you show the error

Segmentation fault

Nothing more

i think you need debug

parse-bbox-func-name=NvDsInferParseCustomTfSSD
custom-lib-path=/home/netcoz/CustomLibPath/libnvds_infercustomparser.so

follow you output

in sgie config i see: network-mode=1 but use parse-bbox-func-name=NvDsInferParseCustomTfSSD. must be parse-bbox-func-name=NvDsInferClassiferParseCustomSoftmax

Note: NvDsInferClassiferParseCustomSoftmax work for resnet so you need custom this function to work with onnx

1 Like

nvdsinfer_customclassifierparser.cpp (4.3 KB)

Found this and added below lines it works now.

parse-bbox-func-name=NvDsInferParseCustomONNX
custom-lib-path=/opt/nvidia/deepstream/deepstream-5.1/sources/libs/nvdsinfer_customparser/libnvds_infercustomparser.so

Thank you @PhongNT for support

1 Like