zbf1991 / RRM

Reliability Does Matter: An End-to-End Weakly Supervised Semantic Segmentation Approach
94 stars 11 forks source link

How can I use my dataset to train the model? #20

Open ST-Jack opened 2 years ago

ST-Jack commented 2 years ago

My dataset only have one class, could you share the way to modify the code?

zbf1991 commented 1 year ago

Please prepare the .npy file to encode your class as a one-hot encoding (you can find a .npy file in this project). Then prepare your train.txt like the "train_aug.txt" in this project. Finally, please prepare your image in VOCdevkit style. After modifying the path as your file path, you can train on your dataset