TextCAVs: Debugging vision models using text
This method reduces the cost and delay in generating explanations for deep learning models, particularly beneficial in domains like medical imaging where labeled data is expensive.
The paper tackles the high cost of obtaining labeled concept examples for concept-based interpretability methods by introducing TextCAVs, which uses vision-language models like CLIP to create concept activation vectors from text descriptions instead of images, reducing annotation effort and enabling interactive debugging.
Concept-based interpretability methods are a popular form of explanation for deep learning models which provide explanations in the form of high-level human interpretable concepts. These methods typically find concept activation vectors (CAVs) using a probe dataset of concept examples. This requires labelled data for these concepts -- an expensive task in the medical domain. We introduce TextCAVs: a novel method which creates CAVs using vision-language models such as CLIP, allowing for explanations to be created solely using text descriptions of the concept, as opposed to image exemplars. This reduced cost in testing concepts allows for many concepts to be tested and for users to interact with the model, testing new ideas as they are thought of, rather than a delay caused by image collection and annotation. In early experimental results, we demonstrate that TextCAVs produces reasonable explanations for a chest x-ray dataset (MIMIC-CXR) and natural images (ImageNet), and that these explanations can be used to debug deep learning-based models.