qtzx06

qtzx06 / yolodex

Public

agent skills for autonomous data labeling, winner at openai codex hackathon 2026

74
11
100% credibility
Found Feb 09, 2026 at 46 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
Python
AI Summary

Yolodex automates turning unlabeled videos into trained object detection models by extracting frames, AI-labeling them, augmenting data, training with YOLO, and iterating until target accuracy.

How It Works

1
🔍 Discover Yolodex

You find a handy tool that turns everyday videos into smart detectors for spotting objects like players or cars without any manual drawing.

2
🛠️ Set it up easily

You run a quick setup on your computer that grabs everything needed to get started in minutes.

3
🎥 Pick your video and targets

You share a YouTube gameplay video link and list simple things to detect, like weapons or vehicles, and set your accuracy goal.

4
⚙️ Start the automatic process

With one command, it pulls out picture frames from the video, smartly marks the objects, creates extra variations, and begins training.

5
🔄 Watch it learn and improve

It keeps checking results, relabeling tricky spots with helper agents, retraining, and looping until it reaches your target accuracy.

6
📊 Review previews and scores

You peek at labeled picture previews, accuracy reports, and even a video recap to see how well it's spotting things.

🏆 Use your custom detector

You get a ready-to-use model file that accurately finds your chosen objects in any new videos or images.

Sign up to see the full architecture

5 more

Sign Up Free

Star Growth

See how this repo grew from 46 to 74 stars Sign Up Free
Repurpose This Repo

Repurpose is a Pro feature

Generate ready-to-use prompts for X threads, LinkedIn posts, blog posts, YouTube scripts, and more -- with full repo context baked in.

Unlock Repurpose
AI-Generated Review

What is yolodex?

Yolodex automates turning unlabeled videos—like YouTube gameplay clips—into trained YOLO object detection models using Python and OpenAI Codex agent skills. Feed it a video URL, target classes (player, weapon, vehicle), and accuracy goal via config.json; it extracts frames, labels with vision AI, augments data, trains, and evaluates in a loop via yolodex.sh CLI until targets hit. You get weights, eval metrics, and label previews ready for deployment.

Why is it gaining traction?

Parallel subagents speed up labeling on big datasets, with dispatch.sh spawning git worktrees for agent github copilot-style parallelism—far faster than manual tools. Supports label modes like GPT-4o, Gemini, or CUA+SAM, plus autonomous loops that iterate smartly. As the 2026 OpenAI Codex hackathon winner, its agent skills github integration draws devs eyeing yolodex ai for agent github repo experiments.

Who should use this?

Game devs labeling objects in footage for detectors. ML engineers prototyping YOLOv8 from scratch videos without annotators. Researchers needing quick baselines on custom classes like UI elements or enemies.

Verdict

Worth forking for video-to-YOLO pipelines if you tolerate early bugs—46 stars and 1.0% credibility signal low maturity, but thorough docs and setup.sh make onboarding fast. Test on small datasets first.

(198 words)

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.