Overview
This course explores the use of gaze data as a supervision source for training deep learning models in medical image classification tasks. The learning outcomes include understanding how gaze data can be leveraged for model training, identifying gaze features containing class discriminative information, and incorporating gaze features into deep learning pipelines. The course teaches methods such as Gaze-WS for weak label extraction and Gaze-MTL for multi-task learning. The teaching method involves a presentation followed by interactive discussions and Q&A sessions. The intended audience includes individuals interested in AI, healthcare applications, and medical image analysis.
Syllabus
Intro
What are observational signals?
Problem settings for observational signals
Gaze-MTL
Hidden stratifications cause model failures
Investigating other observational signals
Domino: An evaluation framework for subgroup robustness
Taught by
Stanford MedAI