Poznan_classification

The goal of the project was to recognize most popular places from Poznań using Bag of Visual Words.

The training dataset was provided by a lecturer. It contained 32 images per class -> 160 images. I've used only those images to train and I haven't used any data augmentation.

Results:

• private test set: 96%
• final, unknown, lecturer test set: 90%

Project requirements

• Python 3.8
• see requirements.txt

Usage

In the train folder you can find jupyter notebook and python script. You can use any of them to train your vocabulary model and classificator. Note that training the vocabulary model might take long.

In the main folder you can find vocabulary model and classificator that was already trained by me. Run main.py to test the model and classificator.