This code was hacked together during Social Impact Data Hackaton 2017 as part of Loxodon challenge within 48 hours.

See also this blog post.

This repository uses (Git Large File Storage)[https://git-lfs.github.com/] for storing trained models. Ensure that the git lfs extension is installed before cloning.

Cars Dataset http://ai.stanford.edu/~jkrause/cars/car_dataset.html

Dataset http://imagenet.stanford.edu/internal/car196/car_ims.tgz Labels http://imagenet.stanford.edu/internal/car196/cars_annos.mat

Other possible sources:

• scraping Google image search

• scraping car sales portals (mobile.de, autoscout24.com, auto24.ee, marktplaats.nl etc)

This project uses pipfile to manage its Python dependencies.

Additionally the app requires that ffmpeg would be installed.

Simple UI with file upload with Django To run Django you need Python (3) and Django (latest) downloaded. When on command line:

$ python manage.py migrate

$ python manage.py runserver

Open web-browser: http://127.0.0.1:8000/

Build a docker image

$ docker build . -t loxodon

Running

$ docker run -it --rm -p 8000:8000 loxodon