MathijsNL

For me it seems to be working to do the following after a disconnect (plugging out and plugging in the network cable): ``` ia.destroy() h.update() ia = h.create() ``` Can...

Not sure if anything has changed in the meantime, but for me PIL is waaaay faster than OpenCV. Python 3.10.6 cv2 4.6.0 PIL version 9.2.0 Using compression level 9, and...

> whereas the original post isn't using compression. That is simply not true. Doing so would result in a file size as big as a BMP image. I modified the...

Looking at the export documentation, int8 is supported by TF Lite and OpenVINO. When trying to export with OpenVINO it is necessary to supply calibration data (not sure about the...

How did you do the conversion? For my datasets the model performance in the test set was unchanged. I used OpenVINO as export format using int8.

I think this Nvidia forum post might answer your question: [https://forums.developer.nvidia.com/t/int8-calibration-in-python-with-tensorrt-8-6/257643/6](https://forums.developer.nvidia.com/t/int8-calibration-in-python-with-tensorrt-8-6/257643/6)

I cloned the repo and used the openvino_int8_cls branch to test, and it seems to work by pointing to the data directory with the image folders. When running the export...

Can you connect to the camera and start acquisition without loading the XML file, or does that also give errors? If so, there might be something wrong with the GenTL...

The container doesn't throw any errors anymore when I add the volume. But I am also not completely sure why the db migration would need the GCP.

Same here! No clue how to fix this.