Superpowers

Automated object detection dataset labeling. Type a prompt, choose an image source, and get a YOLO-labeled dataset — reviewed and uploaded — without manual annotation.

How It Works

1. Acquire images — use existing images, scrape the web (Bing, Baidu, Google, Open Images), or extract frames from YouTube videos
2. Detect objects — Roboflow Universe model discovery + local YOLO inference, with Gemini fallback for uncertain detections
3. Review labels — tinder-style swipe UI with bounding box overlays, undo support, and threshold restart
4. Upload dataset — push curated images + labels directly to a Roboflow project

The pipeline uses three-tier confidence routing: high-confidence detections are kept, uncertain ones get a Gemini second opinion, and low-confidence images are removed. Errors at any stage preserve images — no data loss from transient failures.

Quick Start

# 1. Set up Python environment
python -m venv .venv
source .venv/bin/activate
pip install -e .

# 2. Install frontend dependencies
cd ui && npm install && cd ..

# 3. Set API keys in .env
#    GEMINI_API_KEY=...
#    ROBOFLOW_API_KEY=...

# 4. Launch (starts backend on :8001 + frontend on :5173)
./start.sh

The home screen lets you choose an image source and prompt. For existing images, add them to dataset/images/ before launching.

CLI Usage

Run the labeling pipeline directly without the UI:

python -m detection_pipeline --prompt "basketball"
python -m detection_pipeline --prompt "fire hydrant" --source web --count 200

Options: --threshold, --source (existing/web/youtube), --count, --image-dir, --label-dir, --keep-model-cache, --refresh-model.

Tech Stack

Layer	Technology
CV inference	Roboflow inference SDK (local YOLO)
LLM fallback	Google Gemini API (gemini-2.5-flash)
Model search	Roboflow Universe API
Backend	FastAPI + uvicorn
Frontend	React 19 + TypeScript + Vite 5
Styling	Tailwind CSS 4 + Framer Motion 12
Image scraping	icrawler + CLIP dedup
Video frames	yt-dlp + ffmpeg
Testing	pytest (73 tests across 5 modules)

Project Structure

detection_pipeline/   # Core pipeline: discovery, inference, Gemini, YOLO utils
api/                  # FastAPI review server (keep/discard/undo/upload)
ui/                   # React + Vite review UI (ocean-themed)
tools/                # Image acquisition (web scraper, YouTube frame extractor)
tests/                # 73 unit + integration tests
models/               # YOLO weights (gitignored, auto-downloaded on first run)
dataset/              # images/ and labels/ (gitignored)
docs/                 # Design specs

Tests

pytest tests/