AnyiRao/SceneSeg

Evaluation Metrics

ArtemisiaW opened this issue · 4 comments

Your paper uses 3 metrics, which are AP, Miou and Recall.

I wonder whether the best results of these three metrics are achieved by the same pretrained lgss model, or they are achieved by different pretrained models.

And can you provide the pretrained models?

Hi Dear @ArtemisiaW, thanks for your interest. You may refer to this https://drive.google.com/open?id=1F-uqCKnhtSdQKcDUiL3dRcLOrAxHargz

Are the metrics (recall, AP, miou) achieved by the same checkpoint with the same post-processing or do you select a best checkpoint for each metric?

What the paper report is achieved by the same checkpoint. But I am sure that you could train different models to achieve the best for different metrics.

OK, thank you very much! @AnyiRao