Hello everyone
I am using FaterRCNN network and provide [0.0, 0.0, 0.0, 0.0] boxes with label=0, as a negative example.
I get the error ValueError: All bounding boxes should have positive height and width. Found invalid box [0.0, 0.0, 0.0, 0.0] for target at index 1.
Additional info:
- I don’t have IDs for images;
- I tried passing area to a
targets {}
dictionary; - This did NOT work for me as well: Train Faster R-CNN with negative samples by mnc537 · Pull Request #1911 · pytorch/vision · GitHub
@ptrblck tagging you, because you are some kind of guardian angel of this forum)