protossw512 / AdaptiveWingLoss

[ICCV 2019] Adaptive Wing Loss for Robust Face Alignment via Heatmap Regression - Official Implementation
Apache License 2.0
395 stars 88 forks source link

Does the 'GTBbox' mean the bbox generated by MTCNN or the coordinates in annotation? #3

Closed PlumedSerpent closed 4 years ago

PlumedSerpent commented 4 years ago

Hi, thanks for sharing your implementation. I am kinda confused about the 'GTBbox' in the result table. Does it mean the bbox generated by the MTCNN detector?

protossw512 commented 4 years ago

As shown in the paper, GTBbox means faces are cropped according to ground truth landmarks.

PlumedSerpent commented 4 years ago

How can I switch to GTBbox in your test code?

protossw512 commented 4 years ago

@PlumedSerpent To do this you will need to use min, max of the groundtruth landmarks to find minx, miny, maxx, maxy, instead of reading these from label files. However, I would not recommend you to do this. In real-world applications, you only have bounding boxes from a face detector. If you want to do research, other works also use the detection bounding boxes for benchmarking.