movie-annotation

What problem does it solve?

You need frame-level visual features from movies that map onto EEG-relevant low-level, depth, motion, and semantic information so you can model or analyze how brains respond over time.

Core Features & Use Cases

• Frame-level EEG-relevant feature extraction: Computes low-level luminance/contrast/color/texture, monocular depth, optical-flow motion and scene-cut flags, and face/object counts and categories.
• Semantic scene modeling with CLIP: Generates CLIP embeddings plus zero-shot scene category and naturalness/open-vs-enclosed scores for each frame.
• Reproducible batch pipeline for multiple videos: Runs chunked CPU/GPU inference and writes per-movie outputs for downstream EEG modeling.

Quick Start

Use the skill to annotate all mp4 files in the movies directory by running the project’s main entry point.