Thank you for your outstanding work and willingness to make the code public. I want to use 300w-lp to train my network, but I need to know the face area in the picture. I noticed that there is a roi field in the .mat tag. Is it a face bounding box calibration? If so, there is a negative value, how to restore the coordinates of the face bounding box corresponding to the image?
Thank you for your outstanding work and willingness to make the code public. I want to use 300w-lp to train my network, but I need to know the face area in the picture. I noticed that there is a roi field in the .mat tag. Is it a face bounding box calibration? If so, there is a negative value, how to restore the coordinates of the face bounding box corresponding to the image?