lekoundanmgedeon/food-classifications

Deep learning competition subject to recognize african food classifications # 🍲 Food Image Recognition Challenge ## 📌 Overview This challenge focuses on training a **Deep Learning model capable of recognizing food dishes from images**. Participants must build an **image classification system** that predicts the correct food category for each image. The goal of this project is to practice **Computer Vision with Deep Learning**, especially using: - Convolutional Neural Networks (CNNs) - Transfer Learning - Data augmentation techniques ### Example Classes Some examples of food categories included in the dataset: - Akara - Banga Soup - Massa - Ewedu Soup - Jollof Rice --- # 📂 Dataset We will use the **Nigerian Food Dataset** available on Mendeley: 👉 data.mendeley.com *(This is the dataset to use for the challenge.)* ### Dataset Content The dataset contains: - Images of various Nigerian food dishes - Labels corresponding to each food category - Training and test splits Participants must train their models using the **training images** and generate predictions for the **test images**. --- # 🎯 Learning Objectives By completing this challenge, participants will learn how to: - Build an **image classification pipeline** - Train **Convolutional Neural Networks (CNNs)** - Apply **image preprocessing and data augmentation** - Use **transfer learning** with pretrained models - Compare different model architectures - Evaluate models using appropriate metrics --- # ⚙️ Project Workflow Participants are expected to follow the typical **machine learning workflow**. ## 1️⃣ Download the Dataset Download the dataset and place it inside the `data/` directory. Example: ```text data/ ├── train/ └── test/ ``` --- ## 2️⃣ Data Preprocessing Typical preprocessing steps include: - Resize images (e.g., **224 × 224**) - Normalize pixel values - Convert images to tensors ### Data Augmentation (Recommended) To improve model generalization, apply techniques such as: - Random horizontal/vertical flip - Random ro …