Skip to content

This project modifies the classic VGG16 architecture to classify images into four distinct categories with high accuracy. It incorporates data augmentation, dynamic learning rate adjustments, and comprehensive performance evaluation using accuracy metrics and confusion matrices. Built with PyTorch and supported by a suite of powerful libraries

Notifications You must be signed in to change notification settings

Yonas650/Image-classification-VGG16

Repository files navigation

Image Classification Using VGG16 Model

Project Overview

This project involves developing an image classification system using the VGG16 model, a well-known architecture in the field of deep learning. The objective is to accurately classify images into four predefined categories, showcasing the model's ability to learn and predict from visual data.

Key Features

Customized VGG16 Model: Leveraged the VGG16 model with modifications to the classifier to suit my specific classification needs. Data Augmentation: Implemented several image transformation techniques to enhance model generalization. Dynamic Learning Rate Adjustment: Utilized a learning rate scheduler to optimize the training process. Performance Evaluation: Employed accuracy metrics and a confusion matrix for comprehensive model evaluation.

Technologies Used

Python PyTorch torchvision PIL (Python Imaging Library) NumPy Seaborn Matplotlib scikit-learn

Setup and Usage

  1. Environment Setup:

Ensure Python 3.x is installed. Install required libraries: pip install -r requirements.txt.

  1. Training the Model:

Run python train.py to start the training process. Model checkpoints and logs will be saved in trained.pt

  1. Running Predictions:

Use python test.py to run predictions on new images.

Results and Discussion

Train Accuracy= 100%, Validation Accuracy=90.77%

About

This project modifies the classic VGG16 architecture to classify images into four distinct categories with high accuracy. It incorporates data augmentation, dynamic learning rate adjustments, and comprehensive performance evaluation using accuracy metrics and confusion matrices. Built with PyTorch and supported by a suite of powerful libraries

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages