CAAC: Co-attentive Actionability Classification

This repository contains code for CAAC (Co-attentive Actionability Classification), a transformer-based model that aligns narration and visual frames to predict actionability (PEMAT Guidelines,) in videos.

Structure

caac/ → source code (co-attention, aggregation, entropy, CV=5 testing)

requirements.txt → dependencies

.gitignore → ignored files

Quick start

pip install -r requirements.txt