Food-101 Image Classification with EfficientNet

This project focuses on fine-tuning the EfficientNet-B0 model for image classification using the Food-101 dataset. The model is trained to classify images into 101 different food categories.

Model Summary

Model Architecture: EfficientNet-B0
Dataset: Food-101
Number of Classes: 101
Input Image Size: 224x224
Normalization: Mean - [0.485, 0.456, 0.406], Std - [0.229, 0.224, 0.225]
Optimizer: Adam
Loss Function: CrossEntropyLoss
Learning Rate: 0.001
Epochs: 10

Dataset

The Food-101 dataset consists of 101 food categories, with 1,000 images for each category. The dataset is already split into training (75,750 images) and validation (25,250 images).

Dataset link: Food-101 on Hugging Face

Model Training

The model is based on EfficientNet-B0, a state-of-the-art architecture for image classification. The pretrained model is fine-tuned on the Food-101 dataset using PyTorch.

Training Process

Images are resized to 256x256 and center-cropped to 224x224.
Data augmentation techniques such as random horizontal flip and random rotation were applied to increase the variety of training images.
The Adam optimizer was used with a learning rate of 0.001, and cross-entropy loss was used for classification.

Model Performance

Loss: 1.0926
Validation Accuracy: 0.6944
Test Accuracy: 0.8088

These metrics were achieved after 10 epochs of training

How to use the model

You can use the model directly in your own projects by loading it from hugging face.

Requirements

Python 3.7+
PyTorch 1.9+
torchvision 0.10+
PIL (Pillow)
Hugging Face transformers library

Model Card

ModelType: EfficientNet-B0
Dataset: Food-101
Framework: Pytorch
Optimized for: Image Classification

dwililiya
/

food101-model-classification