Pre-trained and fine-tuned models for computer vision, natural language processing, and medical AI.
Multi-modal VLM for radiology report generation from CT. SOTA Macro F1: 0.429 on CT-RATE.
V2 of Guided LeJEPA with pathology-prioritized sampling and combined auxiliary loss.
Self-supervised LeJEPA model for chest CT with SIGReg regularization, trained from scratch.