search

tianzhi0549 / FCOS

FCOS: Fully Convolutional One-Stage Object Detection (ICCV'19)
https://arxiv.org/abs/1904.01355
Other
3.27k stars 630 forks source link

inference label is wrong #231

Open Amore-HDU opened 4 years ago

Amore-HDU commented 4 years ago

Hello， I have trained FCOS on my own dataset. But when I tried to test my model, I found bounding box was right while label was wrong. I have changed CATEGORIES in demo/predictor.py. @tianzhi0549

Amore-HDU commented 4 years ago

I found when I trained my FCOS, it occured masks = [obj["segmentation"] for obj in anno] KeyError: 'segmentation' so I comment masks = [obj["segmentation"] for obj in anno] masks = SegmentationMask(masks, img.size, mode='poly') target.add_field("masks", masks). I think the reason of wrong label can be it

Mahmood-Hussain commented 3 years ago

I found when I trained my FCOS, it occured masks = [obj["segmentation"] for obj in anno] KeyError: 'segmentation' so I comment masks = [obj["segmentation"] for obj in anno] masks = SegmentationMask(masks, img.size, mode='poly') target.add_field("masks", masks). I think the reason of wrong label can be it

hello did you find what was exactly wrong with your code ? I am also getting masks = [obj["segmentation"] for obj in anno] KeyError: 'segmentation'

Mahmood-Hussain commented 3 years ago

Hello， I have trained FCOS on my own dataset. But when I tried to test my model, I found bounding box was right while label was wrong. I have changed CATEGORIES in demo/predictor.py. @tianzhi0549

My labels are also wrong. Did you get an answer for this? @Amore-HDU