Final project for Data Science bootcamp organized by Sages
Important:
To run this project it is necessary to download dataset from Stanford University page - https://ai.stanford.edu/~jkrause/cars/car_dataset.html.
After downloading tar file of all images (section Download -> Update), please unpack all images and store them in dataset_original/car_ims so the directory looks as follows:
- dataset_original/
-car_ims/
*all images should be stored here*
- cars_annos.mat
cars_annos.mat file is already provided in this repository.
Please use Python 3.9 or higher version. Install libaries from requirements.txt
The project goal is to classify a brand of car based on its image. To build a model, the transfer learning method will be used. Pretrained models that are taken into consideration in this project are as follows:
- ResNet50
- InceptionV3
- MobileNetV2
- VGG19
Moreover, some necessary changes will be applied to top layers of pretrained models to match the requirements of specific dataset. Later on, fine-tuning will be used, to improve models that have the most potential.
Example of prediction on a sample