Scientific Domain Knowledge Improves Vision-Language Fundus Models
The episode discusses the paper 'Scientific Domain Knowledge Improves Vision-Language Fundus Models,' detailing the creation of 'PubMed-Ophtha.' This resource contains 102,023 image-caption pairs from scientific literature. The hosts examine the methodology, which involves a multi-step process using detection models and LLMs to handle complex figures. They conclude by reviewing validation results, including high BLEU scores and mAP metrics.