This project demonstrates image classification using a pre-trained model (VGG16) through transfer learning. The CIFAR-10 dataset is used, which consists of 60,000 32x32 color images in 10 different classes.
In this project, we leverage the power of transfer learning by using the VGG16 model, pre-trained on the ImageNet dataset, to classify images from the CIFAR-10 dataset. The final layers of the model are fine-tuned to fit our classification task.
The CIFAR-10 dataset is used in this project. It includes the following:
- 50,000 training images
- 10,000 test images
- 10 classes: airplane, automobile, bird, cat, deer, dog, frog, horse, ship, truck
- Base Model: VGG16 pre-trained on ImageNet
- Fine-tuned Layers: The last layers of VGG16 are replaced with fully connected layers tailored for the CIFAR-10 classification task.
- Optimizer: Adam
- Loss Function: Categorical Crossentropy
- Resizing images to 32x32 pixels to fit the model's input requirements.
- One-hot encoding of the labels.
- Data augmentation using rotation, zoom, shift, and flip techniques to prevent overfitting.
- Batch size: 64
- Epochs: 25
- Callbacks: Early stopping and model checkpointing are used to save the best model and avoid overfitting.
- Data Augmentation: Applied to enhance the model's ability to generalize.
The model achieved the following performance metrics:
- Training Accuracy:
XX% - Validation Accuracy:
XX% - Training Loss:
XX - Validation Loss:
XX
- Clone the repository:
git clone https://github.com/YourUsername/YourRepoName.git
- Install the dependencies:
pip install -r requirements.txt
To run the model training, use the following command:
python train_model.pyYou can find the saved model and training logs in the models/ directory.
This project demonstrates how transfer learning can be effectively used for image classification tasks. By leveraging pre-trained models, we achieve high accuracy with less computational power and time.
This project is licensed under the MIT License.