AI PROJECT

Car Damage Classifier - Interactive Demo

Upload a vehicle image to get instant AI-powered damage classification with confidence scores

Endpoint:

Drag & drop image here

or click to browse

Best results: clear lighting, close-up of damage, minimal glare
PNG, JPG, JPEG up to 10MB
Ready
Top Prediction
—
Confidence
—

Top-5 Predictions

Upload an image to see predictions
Uses @gradio/client to call /predict_damage endpoint

Supported Classes (12)

D
Car Dent
S
Car Scratch
W
Cracked Windshield
B
Broken Bumper
T
Flat Tire
Fl
Flood Damage
Fi
Fire Damage
H
Hail Damage
M
Broken Side Mirror
R
Rust/Corrosion
V
Vandalism/Keyed
✓
No Damage

How It Works

1. Upload Image

Provide a clear photo of the vehicle damage

2. Preprocess

Image is resized and normalized for the model

3. Run Model

ResNet50 classifies the damage type

4. Return Predictions

Top-5 classes with confidence scores

Use Cases

Insurance Claims

Automate first-pass damage assessment for insurance claim triage and reduce manual review time.

Fleet Inspection

Standardize vehicle condition reporting for fleet management and regular inspection workflows.

Repair Intake

Speed up shop intake processing with instant damage categorization and severity estimation.

Resale Assessment

Document vehicle condition for used car listings, marketplaces, and resale valuation.

Tips & Limitations

Tips for Best Results

  • Use good lighting and avoid glare or reflections
  • Crop close to the damaged area when possible
  • Capture straight-on angles for better clarity
  • Ensure the image is clear and in focus
  • Avoid motion blur or low-resolution images

Limitations

  • Scratches and dents can be visually similar
  • Low confidence (<60%) requires manual review
  • Multiple damage types in one image may confuse the model
  • Subtle damage may be missed in low-contrast images
  • Model trained on specific dataset characteristics