/food-x-251-classification

Classification, Image Preprocessing and Similarity Retrieval on FoodX-251 datset

Primary LanguageJupyter Notebook

FoodX-251 Classification

This repository contains the code and resources for the FoodX-251 project, which aims to design, implement, and evaluate a system or app using the FoodX-251 dataset. The project focuses on fine-grained food classification and includes several components for classification, retrieval, evaluation, and visual analysis.

Project Description

The project is based on the FoodX-251 dataset, and the team is required to develop an original project using the entire dataset or a subset of it. The choice of the subset must be justified. The main components of the project include:

  1. Classification of the validation set into the 251 classes defined in the dataset (fine-grained food classification).
  2. Classification of the degraded validation set into the 251 classes defined in the dataset (fine-grained food classification).
  3. Similarity retrieval (category search) of a query image.
  4. Objective evaluation of the classification and retrieval results.
  5. Visual analysis of significant cases.
  6. Objective comparison of different strategies attempted to achieve the final solution.

Repository Directories

  • Data Cleaning: This directory contains scripts and resources related to data cleaning and preprocessing.
  • annot: This directory stores annotations or additional data for the dataset.
  • notebooks: This directory contains Jupyter notebooks used for experimentation and analysis.
  • scripts: This directory contains scripts and code used for various tasks in the project.
  • README.md: This file, providing an overview of the project and repository.
  • VIPM Presentazione.pptx: A PowerPoint presentation showcasing the project for the VIPM (Very Important Project Members).

Please refer to the respective directories for more detailed information about their contents and usage.

Notes

  • The primary goal is to achieve the highest possible classification performance on the degraded validation set.
  • It is expected that there will be a significant drop in classification accuracy when transitioning from the validation set to the degraded validation set.

For further details, please refer to the project documentation and code within this repository.