In the instance segmentation, one of the image recognition tasks, the indicators mask AP and box AP appear in the paper, but no definition of how to calculate them is found (both Japanese and English).
Could you tell me who knows?Thank you for your cooperation.
machine-learning image
If you specify the "segm" parameter for the code below, it is mask, and if you specify "bbox", it is box.
https://github.com/cocodataset/cocoapi/blob/8c9bcc3cf640524c4c20a9c40e89cb6a2f2fa0e9/PythonAPI/pycocotools/cocoeval.py#L29
Segm is related to mask from the following comments.
https://github.com/cocodataset/cocoapi/blob/8c9bcc3cf640524c4c20a9c40e89cb6a2f2fa0e9/PythonAPI/pycocotools/cocoeval.py#L102
"AP" means average precision.
https://github.com/cocodataset/cocoapi/blob/8c9bcc3cf640524c4c20a9c40e89cb6a2f2fa0e9/PythonAPI/pycocotools/cocoeval.py#L430
© 2024 OneMinuteCode. All rights reserved.