NvOSD_MaskParams Missing Documentation

[MadisonL]

• Hardware Platform (GPU)
• DeepStream Version 6.1
• TensorRT Version 8.2.5-1
• NVIDIA GPU Driver Version 515.43.04/ Cuda 11.7
• Issue Type (missing feature?)

Quick question regarding segmentation models within pyds.

As noted here, obj_meta should contain mask_params as metadata in what I assume is a similar format to rect_params, but “see NvOSD_MaskParams” doesn’t actually link to anything, whereas the rest of the examples on the page do. When attempting to actually use obj_meta.mask_params, we run into the issue:
TypeError: Unable to convert function return value to a Python type! The signature was (self: pyds.NvDsObjectMeta) -> _NvOSD_MaskParams

I’ve seen some other threads on here going back a year with the pybind11 issue, and understand that we may have to recompile with some modifications. My question is, what exact modifications would need to be done to have access to the mask metadata inside of python? Currently we are using a parser almost identical to the maskrcnn parser you guys provide to create a NvDsInferInstanceMaskInfo obj, so my hope would be that it wouldn’t take a lot of change.

Any help would be appreciated.

[yuweiw]

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

You need to bind this stucture by yourself. You can refer the link below:
https://forums.developer.nvidia.com/t/accessing-mask-parameters-from-nvosd-maskparams-in-python-not-using-triton/216724/6
After the code is modified, you need to recompile and replace the library, please refer the link below:
https://github.com/NVIDIA-AI-IOT/deepstream_python_apps/tree/master/bindings