MiDaS on Deepstream

[hardik13]

Hello Everyone, I was trying to run MiDaS( MiDaS/mobile at master · isl-org/MiDaS) on Deepstream. Below is the configuration, I am using

• Hardware Platform (Jetson / GPU) - Nvidia L4
• DeepStream Version - DS 7.1
• JetPack Version (valid for Jetson only) - NA
• TensorRT Version - 10.3.0.26
• CUDA Version - 12.6
• NVIDIA GPU Driver Version (valid for GPU only) - 560.35.05
• Issue Type( questions, new requirements, bugs) - I have created a parser file for postprocessing the output depth map from MiDaS. Deepstream requires a .so file to run. I am not able to convert my .cpp file to .so file using Make. Below is the link to all the required files :- Post-processor - Google Drive

• To reproduce the issue? - The folder should contain Makefile and
.cpp file. Run the command- “make CUDA_VER=12.6”

I am getting the following error. Please help if someone has encountered this or if someone knows what I am doing wrong.

[junshengy]

This is because there is no correct header file, this is just a c/cpp problem, you need to modify yourself, please refer to sample code in /opt/nvidia/deepstream/deepstream/sources/libs/nvdsinfer_customparser

[hardik13]

Hey @junshengy, Thank you for your response. I am new to this and would love if you could tell me what header file should I add. I looked into the documentation and I found most of it are covered in “nvdsinfer_custom_impl.h”.

[junshengy]

Pleaser refer to SDK API documenation NVIDIA DeepStream SDK API Reference: API Modules | NVIDIA Docs

For most cases, it is enough to refer to nvdsinfer_customparser and the corresponding Makefile.

[yingliu]

There is no update from you for a period, assuming this is not an issue anymore. Hence we are closing this topic. If need further support, please open a new one. Thanks