InBiaseD
This is the official code for CoLLAs 2022 paper, "InBiaseD: Inductive Bias Distillation to Improve Generalization and Robustness through Shape-awareness"
- 5
- Python
- MIT
- 4 y ago
This is the official code for CoLLAs 2022 paper, "InBiaseD: Inductive Bias Distillation to Improve Generalization and Robustness through Shape-awareness"
[Nature Medicine] The Limits of Fair Medical Imaging AI In Real-World Generalization
A protocol-induced label leak in PAD-UFES-20: 13 clinical fields are blank as a block for 824 lesions, none malignant. One content-free bit reproduces 79% of the metadata model's discriminative power and matches all 21 clinical fields in multimodal fusion.