CarCheck Egypt — AI-powered used car inspection
# CheKar — AI Car Inspection for Egypt
CheKar uses AI to inspect used cars from photos. Take 8-12 photos of a car, get a graded report showing exterior condition, paint issues, accident signs, and repair cost estimates — all in Egyptian Arabic.
## How It Works
1. **Seller** photographs the car using the guided camera flow
2. **AI** analyzes photos with YOLO damage detection + Qwen vision model
3. **Report** shows a letter grade (A-F), traffic light status per category, and repair costs in EGP
4. **Buyer** knows what they're getting before seeing the car in person
## What It Assesses
| Category | What It Checks |
| ---------- | --------------- |
| حالة الطلاء | Scratches, paint fading, repaint detection |
| حالة الهيكل | Dents, structural deformation, panel gaps |
| الزجاج والإضاءة | Cracked glass, broken headlights/taillights |
| الداخلية | Seat wear, flood indicators, dashboard condition |
| إشارات حوادث | Repaint patterns, panel misalignment, accident history |
| المستندات | Odometer vs wear consistency, year/model verification |
## What It Cannot Assess (at the moment)
- Engine, transmission, brakes, suspension — needs a mechanic
- Electrical systems — needs diagnostic tools
- Undercarriage — needs a lift
- Odometer accuracy — not guaranteed from photos
## Tech Stack
- **AI:** YOLO11 (damage detection) + Qwen3.5-27B-FP8 (vision analysis with cross-checking)
- **Backend:** FastAPI + Huey (SQLite job queue) on a single VPS
- **App:** Flutter (iOS + Android)
- **Language:** Egyptian Arabic first, English secondary
## Project Structure
```text
carcheck/ # Python backend — AI pipeline, API, scoring
pipeline/ # YOLO + Qwen detection and analysis
scoring/ # Letter grade + traffic light grading system
api/ # FastAPI endpoints + Huey worker
report/ # Arabic JSON + PDF report generation
app/ # Flutter mobile app
training/ # YOLO fine-tuning pipeline
scripts/ # VPS setup and ser …