Hello, I noticed the evaluation metrics used here is different from the original WIDER face result page
http://mmlab.ie.cuhk.edu.hk/projects/WIDERFace/WiderFace_Results.html
In original result page, the #1 performance model at the moment is PyramidBox, which achieved 0.887 in the hard set. I'm just curious, if using the evaluation metrics here, what will the average AP of PyramidBox look like, anyone knows?