Marcos Luciano comments

Results 468 comments of


                                            Marcos Luciano

How to get Landmarks in a buffer probe?

There are example codes in https://github.com/marcoslucianops/DeepStream-Yolo-Face https://github.com/marcoslucianops/DeepStream-Yolo-Pose

QuantRepNCSPELAN4Chunk and remove_redundant_qdq_model

I have a different implementation with modelopt. EDIT: The results below is in a custom code for YOLO and YOLOv9-s Enabling the `QuantRepNCSPELAN4Chunk` (with few changes to act as module),...

QuantRepNCSPELAN4Chunk and remove_redundant_qdq_model

I'm using torch quantization. Basically, I rewrote few things in the quantize and calibrate functions to use `max` with `mse` (from the modelopt functions), used `Quant` modules in `mtq.register`, and...

QuantRepNCSPELAN4Chunk and remove_redundant_qdq_model

I won't set it public for now. I'm using in some projects I have. With the final changes, and removing the QuantRepNCSPELAN4 (same of your tests), it's a little fast...

fatal error: NvInferPlugin.h: No such file or directory

are you using the DeepStream docker image?

cannot convert yoloV8 model to int8

I don't have this issue here.

cannot convert yoloV8 model to int8

Try to export the ONNX model without `--dynamic`.

Segmentation fault (core dumped)

Please use the correct exporter file (from `utils`) to export the model.