Adversarial-Deep-Learning / Adversarial-Deep-Learning/code-soup
Visual Perturbation Metrics
Open
good first issue
Implementation
Priority:High
- Dominant language
- Jupyter Notebook
- Stars
- 17
- Forks
- 17
- PR merge metrics
- No merged PRs in 30d
Description
For evasive whitebox or blackbox attacks, the objective of each attack is to fool the model to predict a different class but making it deceptive by making small changes, these changes are measured in distances for Example the L1/L2 Norm of difference.
Implement these metrics
- [x] L1, L2 ... Lk Norm
- [ ] ISSM
- [ ] PSNR
- [x] SAM
- [x] SRE
You can find numpy and cv2 implementation at https://github.com/up42/image-similarity-measures/blob/master/image_similarity_measures/quality_metrics.py
Contributor guide
Assessment
This issue has not been assessed yet.