This project involves the use of a VGG-16 neural network to distinguish between images of cats and dogs. Built with TensorFlow, it demonstrates effective data preprocessing, fine-tuning a pre-trained model, and evaluating performance with a detailed confusion matrix.
- Objective: Binary classification to differentiate cat and dog images.
- Techniques Used:
- Data augmentation and preprocessing with
ImageDataGenerator. - Transfer learning with VGG-16.
- Model evaluation using a confusion matrix.
- Data augmentation and preprocessing with
The model demonstrates robust performance with a 97.5% accuracy on the validation set during training. Upon evaluation, it achieves a commendable 96.50% accuracy on the test set, indicating strong generalization and predictive capabilities.
The dataset utilized for training and evaluating the model is the popular "Dogs vs. Cats" dataset, which was originally used for a Kaggle competition. It comprises images of cats and dogs, aimed at developing algorithms to correctly classify the images into the respective categories. The dataset can be accessed at Kaggle: Dogs vs. Cats.
This project was inspired by the "Keras with TensorFlow Course - Python Deep Learning and Neural Networks" provided by freeCodeCamp.org. The concepts and code examples from the YouTube tutorial were instrumental in the development of this project, offering a solid foundation in deep learning practices and model evaluation techniques.
Feel free to fork the project and submit pull requests.