jbeomlee93 / AdvCAM

Anti-Adversarially Manipulated Attributions for Weakly and Semi-Supervised Semantic Segmentation (CVPR 2021)
MIT License
128 stars 16 forks source link
advcam cvpr2021 weakly-supervised-learning weakly-supervised-segmentation

Anti-Adversarially Manipulated Attributions for Weakly and Semi-Supervised Semantic Segmentation

Input Image Initial CAM Successive Maps with adversarial climbing
a b c

The implementation of Anti-Adversarially Manipulated Attributions for Weakly and Semi-Supervised Semantic Segmentation, Jungbeom Lee, Eunji Kim, and Sungroh Yoon, CVPR 2021. [paper]

Installation

Step 1. Prepare Dataset

Step 2. Prepare pre-trained classifier

Step 3. Obtain the pseudo ground-truth masks for PASCAL VOC train_aug images and evaluate them

bash get_mask_quality.sh

Step 4. Train a semantic segmentation network

Acknowledgment

This code is heavily borrowed from IRN, thanks jiwoon-ahn!