Hi there,
So I’m working on a Traffic light SSD with focal loss detector, but I have a problem with the bboxes that are get_annotations() return.
For example here is the correct from of one of the bounding boxes [] and here is how it looks in the COCO flavored json file [733, 187, 709, 135] and here is the result I get from the get_annotations method [187, 733, 322, 1442].
I went into the source to see what the method does and found that its doing what its supposed to do append([bb[1],bb[0], bb[3]+bb[1], bb[2]+bb[0]]) my question is why is it doing so, and why did it work with the PASCAL data and not my own?