Sovit Ranjan Rath

Hello @AhmadJamal01 This is most probably happening because after rescaling some of the bounding boxes have negative dimensions. It will be difficult to right away tell which file causes it...

@AhmadJamal01 I meant rescaling.

@AhmadJamal01 Thanks. Can you also add the dataset link without removing the problematic images? Maybe I can reproduce the results and write a script to discard such images on the...

Thanks.

I am working on this. Will post a solution soon.

If you can provide dummy data, it will be easier for me to create a solution for this. Thanks.

Hello, I have just pushed an update to `datasets.py` that removes all files with invalid bounding boxes before training. Please check. It also prints the erroneous file names if needed...

Hello. If you execute `eval.py`, at the end it will print how much time it took for the evaluation. If you are asking just for inference, it will print the...

Follow up. Just noticed that `eval.py` does not print the time. The `inference_video.py` does for each frame. Also, the time is printed after each validation epoch during training.

Hi @unrue You need to execute the training from within the `fasterrcnn-pytorch-training-pipeline` directory.