Semi-Supervised Semantic Segmentation With Region Relevance

Requirements

📋 The code is developed using Python 3.6 with PyTorch 1.10.1. The code is developed and tested using 2 NVIDIA TITAN RTX GPUs.

Training && Evaluation

To train and eval the model in the paper, run this command:

cd ./voc/voc8.res50v3+.CPS+CutMix
bash script_mod.sh

📋 In script_mod.sh, you need to specify some variables, such as the path to your data dir, the path to your snapshot dir that stores checkpoints, etc.

Our Models DownLoad && Test

We have released the weight models on PASCAL VOC in our experiment, you can download here:

• My voc2 model trained on PASCAL VOC 2012 at 1/2 partition protocol.
• My voc4 model trained on PASCAL VOC 2012 at 1/4 partition protocol.
• My voc8 model trained on PASCAL VOC 2012 at 1/8 partition protocol.
• My voc16 model trained on PASCAL VOC 2012 at 1/16 partition protocol.

📋 The command for test is in script_mod.sh, you only need to annotate the command for training and modify the parameters appropriately.