CARZero: Cross-Attention Alignment for Radiology Zero-Shot Classification
Date:
Presented a poster on CARZero, a cross-attention alignment framework for radiology zero-shot classification, at CVPR 2024 in Seattle, USA (Jun 16–22, 2024).
Abstract: CARZero proposes a learnable cross-attention based image-text alignment strategy for zero-shot medical image classification. By replacing conventional cosine similarity metrics with a learnable alignment module, CARZero models complex semantic relationships between radiology images and clinical text, achieving state-of-the-art performance on chest X-ray zero-shot diagnosis benchmarks.