evaluate

[ccxietao]

evaluate

[aby2s]

Hi.
Sorry, I don't exactly understand what you're asking for. It's possible to either evaluate model on COCO validation set or to obtain mask for any image. It's all described in the readme.

[ccxietao]

Hello, I want to know how to use the evaluation index AP of the coco dataset to evaluate the quality of the model. You use the IOU evaluation index in your code, so how should I modify it?

[aby2s]

The best way to do it is to store predictions in COCO format and use official scripts for evaluation. Unfortunately, I didn't implement it, but it's not hard to do.
You can modify the _run_validation method or add a similar one, which calculates predictions (self.score_predictions/self.refinement_prediction for SharpMask) in the session run instead of metric calculation. Then store predictions in COCO format and use COCO evaluation tools to calculate AP - https://github.com/cocodataset/cocoapi/blob/master/PythonAPI/pycocotools/cocoeval.py

[ccxietao]

Ok, I will try it, thank you very much.