self.roirotate.apply(feature_map, rois, self.pooled_height, pooled_width, self.spatial_scale) return tensor with all element is 0

jiangxiluning / FOTS.PyTorch

FOTS Pytorch Implementation

BSD 3-Clause "New" or "Revised" License

643 stars 194 forks source link

self.roirotate.apply(feature_map, rois, self.pooled_height, pooled_width, self.spatial_scale) return tensor with all element is 0 #91

Open tranducanhbk opened 2 years ago

tranducanhbk commented 2 years ago

Hi, Can u show some result of recognize part. I train with my custom data and the detection part work well but when i try to get recognize result it is not as expected. I found the rroiallign retrun a tensor with all element is 0. Can u help me find the rootcause of this issue, Many Thanks

jiangxiluning commented 2 years ago

this bug is fixed now.