FoodVision Big is an advanced image classification project built using PyTorch. This repository showcases the end-to-end workflow of creating, training, and deploying a machine learning model for food image classification.
The project leverages a pre-trained EfficientNet B2 model fine-tuned on the Food101 dataset to achieve high accuracy. It also includes steps to turn the trained model into a deployable application.
| Features | Description |
|---|---|
| EfficientNet B2 Fine-Tuning | Using transfer learning for efficient and accurate image classification. |
| Custom Training Pipeline | Implementation of a modular training process for flexibility and scalability. |
| Model Deployment | Structured as a web app for real-world testing and interaction. |
| Pre-trained Model Management | Save, load, and evaluate the trained model efficiently. |
| App Structure | Guidelines and scripts to deploy the model as a web application. |
| Directory/File | Description |
|---|---|
models/ |
Directory for saving trained models. |
demos/foodvision_big/ |
Application files for deployment. |
09_pretrained_effnetb2_feature_extractor_food101_20_percent.pth |
Pre-trained model weights. |
app.py |
Main app script for deployment. |
class_names.txt |
Class names for Food101 dataset. |
examples/ |
Example images for testing. |
model.py |
Model definition and architecture. |
requirements.txt |
Dependencies for the application. |
| Objective | Details |
|---|---|
| Train a pre-trained EfficientNet B2 model | Fine-tune for the Food101 dataset. |
| Save and load the trained model efficiently | Ensure model compatibility across platforms. |
| Deploy the model as a web application | Allow real-world usability and interaction. |
| Evaluate and optimize the model size | Enhance deployment performance. |
- Clone this repository:
git clone https://github.com/yourusername/foodvision-big.git- Install dependencies:
pip install -r requirements.txt- Train the model or load the pre-trained weights.
- Deploy the app locally:
python app.py
This project is licensed under the MIT License.