This webapp provides a user-friendly interface for the classifiers and visualization tool developed for the car damage assessment capstone project from VKB during the data incubator.
- Python 3 (available from the command line)
- pip
- Git (available from the command line)
git clone https://github.com/gaetjen/capstone_webapp.git
cd capstone_webapp
python3 -m venv app_env
. app_env/bin/activate
pip install -r requirements.txt
export FLASK_APP=app.py
python download_models.py
git clone https://github.com/gaetjen/capstone_webapp.git
cd capstone_webapp
python -m venv app_env
app_env\Scripts\activate
python -m pip install -r requirements.txt
set FLASK_APP=app.py
python download_models.py
If you have a system set up to use tensorflow on the gpu the following may lead to speedups in classification:
pip uninstall tensorflow
pip install tensorflow-gpu
You can use flask run
to start the app and access it at the web address http://127.0.0.1:5000/. Alternatively you can use python app.py
and the site will be available from your own computer at http://0.0.0.0:33507/ and also from other computers via your public IP address.
On the main site there is a form to upload images. After upload and classification is complete the images are shown together with the prediction and confidence, as well as a visualization of the image regions relevant for the damage/no-damage classifier.
The app is not set up to work properly with multiple users in parallel.