FCNs in the Wild Pixel-level Adversarial and Constraint-based Adaptation

To be finished later

Pytorch implemention of this arxiv paper

The FCN model used is papre Multi-scale context aggregation by dilated convolutions note not finished

dataset

• GTA5 dataset

requirements

• tqdm
• pytorch
• numpy
• scipy
• Pillow
• visdom

training

examples

note

In the GTA5 dataset, the label file is png format which uses palette, so to train the model should record the palette infomation to recover the output with color