The quantity of annotation in image

yhenon / pytorch-retinanet

Pytorch implementation of RetinaNet object detection.

Apache License 2.0

2.15k stars 665 forks source link

The quantity of annotation in image #230

Open TowoC opened 3 years ago

TowoC commented 3 years ago

Hi, yhenon. As the sample of readme, I found that the quantity of annotation of csv dataset in each image is just one. And I also found that in dataloader.py. The annotations of data only input one annotation.

I have no idea why the annotation only gets one as the input of retinanet.

And I've revised the code to make it can have multi-annotation as input.

I am just curious about why the original code could only input one annotation in each image.

Daheer commented 1 year ago

Hi Towo, I realized this issue after training for about 8 hours too. Could you share the changes you made to make it have multi-annotation? It would really help me out. Thanks