Is there a bug in dump_tensorflow_weights.py for the first box predictor layer ?
seedlingfl opened this issue · comments
Only the first box's scale is changed from 0.1 to 0.2 when switching from tf to caffe. But the script change the weights of all 3 boxes. Is this a bug?
@seedlingfl I think you have mentioned correctly. I have modified the code to adjust the box only at the first box (new_weights[0,:2]), then my converted caffe model predicts bounding boxes more similar to the tensorflow checkpoint. I think it is highly a bug here!
@seedlingfl I think you have mentioned correctly. I have modified the code to adjust the box only at the first box (new_weights[0,:2]), then my converted caffe model predicts bounding boxes more similar to the tensorflow checkpoint. I think it is highly a bug here!
hello, can you tell me how to modify to fix this problem?