I am trying to run a simple evaluation:
cocoGt = COCO(json_gt_path)
cocoDt = cocoGt.loadRes(json_pr_path)
cocoEval = COCOeval(cocoGt, cocoDt, "bbox")
cocoEval.evaluate()
cocoEval.accumulate()
cocoEval.summarize()
, where my detector did not detect any box-annotations, so json_pr_path only has an empty array [].
This leads to an error in coco.loadRes:
pycocotools/coco.py", line 335, in loadRes if 'caption' in anns[0]: ~~~~^^^ IndexError: list index out of range
Is there a way to use these functions and evaluate with zero predicted/detected boxes?