Scape

Hi, thanks for detailed descriptions of training steps! I want to know if it's necessary to jointly train edge and inpainting models? ( line 164 in edge_connect.py ) Have you...

edge model does not converge

8

Hi, thanks for releasing codes. I am using this code to train my own data from scratch . After I changed the sigma of canny calculator to 1, the feature...

GTA->Cityscapes

1

Thanks for releasing the code! Could you release the GTA->Cityscapes pretrained model? Have you ever tested the fid of this results?

我单独把getSimilarityTransform的程序来出来跑了一下，发现得到的rotation是 shape_to * rotation = shape_from 从代码里来看，注释写的是 // get the rotation and scale parameters by transferring shape_from to shape_to, shape_to = M*shape_from 但是regressor.cpp 108这一段： regression_targets[i] = ProjectShape(augmented_ground_truth_shapes[i], augmented_bboxes[i]) - ProjectShape(augmented_current_shapes[i],...

Inconsistency between paper and code

2

Thanks for releasing codes! I found several differences between the paper and the released code: 1. The batch size. In your paper, the batch size of segmentation model was set...

M(0) model

2

Hi, could you release the M(0) model of DeepLab for GTA5 and SYNTHIA ?

Wrong data preprocessing in demo.py

1

The preprocessing is not aligned with your training in your demo code. For the data preprocessing in training, the face images are first cropped to squares. It seems that u...