How to get the bounding box?

[panda-lab]

Hi,
Thanks for your work. Recently i test your code on 300W test set, include helen, aflw, ibug. when i input the test images with face bounding box based on mtcnn, i can not get a high accuracy. So could you tell me which detector do you adopt in this project. Thank You.
Siyuan

[D-X-Y]

Hi,

I use https://github.com/D-X-Y/SAN/blob/master/cache_data/generate_300W.py#L38 to generate the bounding box on 300W. Specifically, two kinds of the bounding box are used. One is "OD", it uses the official bounding box provided by the 300W website (https://ibug.doc.ic.ac.uk/media/uploads/competitions/bounding_boxes.zip). The other is "GT", use the tight bounding box around the facial landmarks.
I'm not surprised that using other bounding boxes would decrease the performance. Because our framework is not designed for the robustness w.r.t. different detectors. For our SAN, the detector used for training should be the same as the detector used for test.

[CHELSEA234]

Hi @D-X-Y 👏👏👏:

I am new to this topic, maybe some questions sound silly, thanks for your patience and guidance 👍👍👍👍👍.

I have seen your answer in #issue 14. How did you get the bounding box in link? Is this the “GT”, tight bounding box? It looks like the predefined bounding box imported from mat data.

Now I need to execute your code on my own image (single), how should I locate the tight bounding box, can you share the instruction link if exists? or can you tell me how did you do on the bounding box of ../cache_data/cache/test_1.png, that looks pretty good.

Best,
XG

[cmburgul]

The facial bounding box format followed by 300W dataset is ( top-left x, top-left y, bottom right x, bottom right y). The other formats like MTCNN follows (top-left x, top-left y, width, height). I took a out of dataset image and used MTCNN to detect bounding box and changed the format and I am getting facial landmarks ish accurate.