MhLiao / TextBoxes_plusplus

TextBoxes++: A Single-Shot Oriented Scene Text Detector
Other
954 stars 279 forks source link

why 12 variances to be predicted? #126

Open basaltzhang opened 5 years ago

basaltzhang commented 5 years ago

x1, y1, x2, y2, x3, y3, x4, y4 are 8 variances, and the former the xmin, ymin, xmax, ymax can be computed by these 8 variances, so why 12 variances are predicted instead of 8? Have you tested the 8 variances? How does it perform?