Typical Food Classification Using AlexNet, ResNet with Grad-CAM
This project focuses on classifying images of 10 popular food categories using CNN models based on AlexNet and ResNet18, incorporating data augmentation techniques and cyclical learning rate.
This project focuses on classifying images of 10 popular food categories using a dataset containing 10,000 images. The dataset features diverse food types, including apple pie, chicken wings, dumplings, and more.
Technical Approach
- CNN Models: Implemented and evaluated CNN models based on AlexNet and ResNet18 architectures
- Data Augmentation: Applied various data augmentation techniques to enhance training performance and model generalization
- Learning Rate Optimization: Utilized cyclical learning rate scheduling to improve training efficiency
- Visualization: Integrated Grad-CAM (Gradient-weighted Class Activation Mapping) for model interpretability
Dataset
The project uses a comprehensive dataset of 10,000 images across 10 food categories:
- Apple pie
- Chicken wings
- Dumplings
- And 7 other popular food types
Results
The implementation demonstrates effective food classification with improved performance through data augmentation and optimized training strategies. The Grad-CAM integration provides valuable insights into the model’s decision-making process.