[DMP 2026] - Planet Read - End to End Implementation of Intelligent CC Generation (Modules 1,2 and 3)

[Aditi2k5]

This PR expands the project documentation to fully cover the complete intelligent closed caption suggestion pipeline from all three modules through final output generation. The README now explains how the system works end to end, why each module exists, and how the modules interact to produce final subtitle suggestions.

Demo Video Link : https://youtu.be/0BDAB2LCOjs

What’s included

• Full explanation of Module 1 in sound_detector.py, including audio extraction, YAMNet inference, class filtering, category mapping, boosting, merging, and event selection.
• Full explanation of Module 2 in visual_detector.py, including frame sampling, MediaPipe Face Mesh processing, EAR/MAR/brow-based reaction scoring, and temporal aggregation.
• Full explanation of Module 3 in fusion_engine.py, including weighted fusion, priority thresholds, deduplication, SRT gap enforcement, and optional frame annotation.
• End-to-end flow from input video to final output.srt, report.json, and annotated frames.
• A Mermaid flow diagram showing the complete pipeline.
• A full directory structure snapshot of the workspace.
• Setup and run instructions.
• Tech stack, features, advantages, limitations, future work, and troubleshooting notes.

Impact

• Makes the project easier to understand for reviewers, collaborators, and evaluators.
• Provides a clear technical narrative for the full pipeline instead of only isolated module details.
• Documents the current implementation accurately, including outputs and configurable parameters.

[Aditi2k5]

Hello @abinash-sketch,
I have got a complete understanding of the problem we are working to resolve and the vision along with future work. I hope to discuss the project scope, your expectations, how we may work on this for DMP if possible and further. I'm free to discuss regarding the same through a brief call or chat through my discord id, adi._05.