Project of Vehicle type recognition done by Violaine Courrier, Elise Barrère, Carlos Cortés and Paul Fayard, and supervised by Olivier Grisel and Charles Ollion.
The goal of this project is to be able to recognize several types of vehicles (motos, cars, trucks) based on their pictures, using deep learning models. To go further, we have also tried to recognize the brands of these vehicles.
We can see several applications : make market studies of vehicle brands by region, or evaluate the pollution on a certain road. Our project uses pre-trained CNN like VGG-16 and is applied to pictures, but to go further, it would be interesting to use real-time model that we could apply to videos.
The files scrapping.py and scrapping.ipynb explained how do we constructed our proper dataset, with the website tucarro, colombian website to sell and buy vehicles.
The file vehicle_recognition/vehicle_recognition.ipynb explained the models we use to distinguish cars, motos and trucks.
The file vehicle_recognition/deep_dream.ipynb explained what happens concretely in our deep learning models, what features do we learn.
The file brand-recognition/brand-recognition.ipynb explained the models we use to distinguish the different car's brands.