Deepstream 4.0 custom parser function

Hi,
I am using detectnet with deepstream 4.0, I tried with https://github.com/AastaNV/DeepStream, but I get following error

nvinfer gstnvinfer.cpp:511:gst_nvinfer_logger:<primary-nvinference-engine> NvDsInferContext[UID 1]:fillDetectionOutput(): Failed to parse bboxes using custom parse function
Segmentation fault (core dumped)

. How can I solve it? Is there any document for custom parser in deepstream 4.0 .

Hi,

The GitHub is used for JetPack3.2 and deepstream-1.5.
Please use the samples located in deepstream 4.0 instead.

Thanks.

Hi,

I couldn’t find any face detection sample. I am trying to implement using caffe model with custom parser fucntion. I am newbie for deepstream as well as c++. Is there any documentation to follow to create custom parser fucntion? or can you suggest me any samples to follow?

Hi,

You can find our deepstream document here:
https://docs.nvidia.com/metropolis/deepstream/4.0/dev-guide/index.html

To use a customized model, please refer to “Application Customization” section.
In general, just update the file path in the configure to yours.

If your model is different from the resnet output format, you will also need to write a custom bbox parser for it.
There are some samples for your reference:
/opt/nvidia/deepstream/deepstream-4.0/sources/objectDetector_FasterRCNN/nvdsinfer_custom_impl_fasterRCNN/nvdsparsebbox_fasterRCNN.cpp
/opt/nvidia/deepstream/deepstream-4.0/sources/objectDetector_SSD/nvdsinfer_custom_impl_ssd/nvdsparsebbox_ssd.cpp
/opt/nvidia/deepstream/deepstream-4.0/sources/objectDetector_Yolo/nvdsinfer_custom_impl_Yolo/nvdsparsebbox_Yolo.cpp

Thanks.

Hi, I have the same issue. I am trying to apply detectnet in deepstream 4.0 and couldn’t find the parser for it. The only one available at https://github.com/AastaNV/DeepStream was outdated. Can you please update it?

Thanks.