DamageDoctor

Set-Up

Note: Set-up was done with python 3.12.7 version General instructions to set up the environment for all experiments

1. Create venv (virtual environment) with python -m venv venv
2. Activate venv with source venv/bin/activate
3. Install requirements with pip install -r requirements.txt
4. Create a directory named images/, all experiments will use this directory to inference on

• Images of your choice can be added here, but here is one neat dataset from kaggle. Download this and move the images into your images/ dir.

Notes:

• only .jpg and .jpeg files have been tested to work with it
• deactivate venv with deactivate command

Experiments

damage_metric_experiment [VOID]

Ran experiment from article in damage_metric_experiment

• Detects damage in car images using MobileNetV2
• Dataset used: Kaggle Link

img_classification_experiment

Running regular image classification model locally for offline use. Specs:

• MobileNetV3 implemented with PyTorch
• Using large weight file (.pth) trained on ImageNet
• See README.md in directory for additional information to run