Traffy Fondue is an application for people to report urban problems by submitting images and their categories. The aims of this project is to create a image classification model. The dataset containing 17,922 images and 10 categories including sanitary, sewer , stray, canal, light, flooding, electric, traffic , road, and sidewalk.
I preprocessed datasets and then trained with a Vision transformer (ViT) model. The result shows that Vit achieved an F1 score of 81% on the private leaderboard.
You can see the final report of the project in this link. and the code is in the kaggle notebook.