Automatic Concept Extraction for Concept Bottleneck-based Video Classification

Jeya Vikranth Jeyakumar, Luke Dickens, Yu-Hsi Cheng, Joseph Noor, Luis Antonio Garcia, Diego Ramirez Echavarria, Alessandra Russo, Lance M. Kaplan, Mani Srivastava

Keywords: Explainable AI, Video Classification, Concept Extraction

Abstract: Recent efforts in interpretable deep learning models have shown that concept-based explanation methods achieve competitive accuracy with standard end-to-end models and enable reasoning and intervention about extracted high-level visual concepts from images, e.g., identifying the wing color and beak length for bird-species classification. However, these concept bottleneck models rely on a domain expert providing a necessary and sufficient set of concepts--which is intractable for complex tasks such as video classification. For complex tasks, the labels and the relationship between visual elements span many frames, e.g., identifying a bird flying or catching prey--necessitating concepts with various levels of abstraction. To this end, we present CoDEx, an automatic Concept Discovery and Extraction module that rigorously composes a necessary and sufficient set of concept abstractions for concept-based video classification. CoDEx identifies a rich set of complex concept abstractions from natural language explanations of videos--obviating the need to predefine the amorphous set of concepts. To demonstrate our method’s viability, we construct two new public datasets that combine existing complex video classification datasets with short, crowd-sourced natural language explanations for their labels. Our method elicits inherent complex concept abstractions in natural language to generalize concept-bottleneck methods to complex tasks.

One-sentence Summary: A method that extracts a rich set of complex concepts from natural language explanations for concept-based video classification