https://github.com/momo1986/robust_sam_iv

Adversarial study on SAM for autonomous driving
https://github.com/momo1986/robust_sam_iv

Last synced: about 1 month ago
JSON representation

Adversarial study on SAM for autonomous driving

• Host: GitHub
• URL: https://github.com/momo1986/robust_sam_iv
• Owner: momo1986
• Created: 2024-06-20T09:32:04.000Z (7 months ago)
• Default Branch: main
• Last Pushed: 2024-11-04T16:34:23.000Z (2 months ago)
• Last Synced: 2024-11-04T17:27:29.755Z (2 months ago)
• Language: Python
• Homepage:
• Size: 66.4 KB
• Stars: 6
• Watchers: 2
• Forks: 1
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

• Awesome-Segment-Anything - [code

README

        
This repo is an enhanced repo of https://github.com/fudan-zvg/Semantic-Segment-Anything.

The new repo focuses on adversarial robustness of SAM on autonomous driving.

Running on cityscapes:

```bash

python scripts/main_ssa.py --ckpt_path ./ckp/sam_vit_h_4b8939.pth --save_img --world_size 1 --dataset cityscapes --data_dir data/cityscapes/leftImg8bit/val/ --gt_path data/cityscapes/gtFine/val/ --out_dir output_cityscapes_adversarial

python scripts/evaluation.py --gt_path data/cityscapes/gtFine/val/ --result_path output_cityscapes_adversarial/ --dataset cityscapes

```

Run on black-box attacks, please comment the code related to the white-box attacks and add corruptions in the pipeline:

```bash

pip install https://github.com/bethgelab/imagecorruptions

```

Our up-to-date demo is the pipeline under the PGD attacks.