SAAFQ: Scale-Aware Adaptive Feature Quantization for Robust Medical Image Representation Learning

ECAI 2025 (supplementary/code release placeholder)

This repository will host the official implementation, pretrained models, and reproducible scripts for SAAFQ.
Status: Code and models will be released soon after camera-ready. Watch ★ the repo for updates.

---

🧩 Overview

SAAFQ introduces scale-aware adaptive feature quantization to improve robustness of medical image representations under domain shift and perturbations.
Core ideas:

• Adaptive quantization driven by multi-scale statistics
• Robust SSL pretraining compatible with standard backbones
• Strong downstream performance on classification/segmentation

---

🔗 Resources

• Paper: to appear at ECAI 2025
• Supplementary (Zenodo, cite all versions): https://doi.org/10.5281/zenodo.16924156

---