Skip to content

GETA_OS

GETA_OS is a high-performance, automated pipeline designed to generate high-quality AI training datasets from raw video data.

Code Video Cerebral Valley

Core trick

Gemini 2.5 Flash filters video frames for objects of interest, and YOLOv8 supplies bounding boxes that become training labels.

Tags

visiondev tools

Description

GETA_OS is a high-performance, automated pipeline designed to generate high-quality AI training datasets from raw video data. By combining the reasoning power of Gemini 2.5 Flash with the precise localization of YOLOv8, GETA_OS extracts, filters, and labels specific objects of interest to create "plug-and-play" training sets.
ByungyeonKim/gemini-hackathon★ 10 forksPythonCreated Feb 28, 2026 · last push Feb 28, 2026

README excerpt

<div align="center">

# GETA_OS  
### Gemini-Enhanced Training Algorithm  

<p><strong>Transforming Raw Gameplay Video into Structured, Training-Ready Vision Datasets</strong></p>

<br/>

<img src="https://github.com/user-attachments/assets/d4de6473-57cd-4d2d-84bd-1fac944ba9e2" 
     alt="GETA_OS System Preview" 
     width="850"/>

<br/>

<p>
  <em>Automated Frame Extraction • Semantic Filtering • Auto-Annotation • One-Click Dataset Export</em>
</p>

</div>
GETA_OS is an automated pipeline that transforms unstructured video (e.g., gaming livestreams) into training-ready computer vision datasets. By combining multimodal reasoning from Gemini with automatic object localization from YOLO, GETA_OS filters, annotates, and exports domain-specific datasets ready for model fine-tuning.

## 🎯 Why GETA_OS?

Gaming videos contain:
- Complex motion
- Rare lighting/weather conditions
- Synthetic but realistic environments
- High object density
- Unique camera angles

These properties make them a powerful — and underutilized — source of training data for vision models.

GETA_OS unlocks that potential!

## ⚡ Core Features
### 🎯 Dynamic Object Targeting

Specify any detectable object (e.g., car, person, traffic light) to generate a focused dataset.

### 🧠 Gemini Relevance Filtering

Uses Gemini 2.5 Flash to:
- Analyze extracted frames
- Determine semantic relevance
- Remove low-signal or irrelevant frames
  
This reduces noise before labeling.

### 🏷 YOLO Auto-Annotation

Uses YOLOv8 to:

-

Team