Please provide complete information as applicable to your setup.
• Hardware Platform (Jetson / GPU) GPU
• DeepStream Version 6.3
• Issue Type( questions, new requirements, bugs) questions
Hi, I tried to write a parser for the YOLOv8 classification model, and the model works well. The problem is in the preprocessing steps: the YOLOv8-cls model uses CenterCrop, which is not supported in DeepStream. Could you tell me how I can modify or develop this feature?
Noticing CenterCrop is a kind of ROI, nvdspreprecess alreeady support ROI. please refer to sample deepstream-preprocess-test.
Thanks for your advice. However, in my use case, I need more preprocessing steps than just CenterCrop (which is one of them), so I think I should modify the gst-nvinfer plugin. Could you tell me if it’s possible to do this, and which function, step, or file I should consider modifying?
nvinfer only supports regular preprocessing, like scaling and normalization. nvdspreprocess support ROI and customized preprocessing. so we suggest using nvdsprecess+nvinfer solution. for example, deepstream-3d-action-recognition 's preprocess will wrap a sequence of tensors. could you elaborate on your custom preprocessing? gst-nvinfer is opensource. nvinfer’s scaling is in gst_nvinfer_process_full_frame of \opt\nvidia\deepstream\deepstream\sources\gst-plugins\gst-nvinfer\gstnvinfer.cpp. nvinfer’s normalization is on InferPreprocessor::transform of \opt\nvidia\deepstream\deepstream\sources\libs\nvdsinfer\nvdsinfer_context_impl.cpp
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
This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.