Chanyu Moon | 문찬유

About me

I am master student in DGIST, advised by professor Ji-Woong Choi.
I received by Bachelor’s degree in Computer Science and Electronic Engineering from DGIST in 2024.
Check out my cv / github / scholar

Research

I am interested in the brain. I want to understand how our brain performs cognitive functions and motor control, and why patients with brain disorders such as Parkinson’s disease experience certain symptoms.
* states the Equal Contribution

OFF-CLIP: Improving Normal Detection Confidence in Radiology CLIP with Simple Off-Diagonal Term Auto-Adjustment
Junhyun Park*, Chanyu Moon*, Donghwan Lee, Kyungsu Kim, Minho Hwang
arXiv, 2025
arXiv

Contrastive Language-Image Pre-Training (CLIP) has enabled zero-shot classification in radiology. However, conventional contrastive learning struggles with normal case detection due to its strict intra-sample alignment. To address this, we propose OFF-CLIP, a contrastive learning refinement that improves normal detection by introducing an off-diagonal term loss to enhance normal sample clustering and by implementing sentence-level text filtering to mitigate false negatives by removing misaligned normal statements from abnormal reports. Additionally, OFF-CLIP can be applied to radiology CLIP models without requiring any architectural modifications. Compared to the current state-of-the-art model, CARZero, OFF-CLIP significantly improves normal classification (0.61 AUC increase on VinDr-CXR). Furthermore, our model enhances zero-shot grounding by improving pointing game accuracy.