Convolutional Neural Networks and Computer Vision
Muhammad Haseeb
Food Classification with CNN and CV
The project "CNN and Computer Vision for Image Classification" leverages the power of Convolutional Neural Networks (CNNs) to distinguish between different categories of images, specifically pizza and steak. Utilizing TensorFlow and Keras, the project applies various deep learning techniques to accurately classify images.
The workflow starts with the acquisition of the dataset, which includes images of pizzas and steaks. These images are then preprocessed for use in the neural network models. The preprocessing steps include resizing the images to a uniform size and normalizing the pixel values.
Several CNN architectures are tested within the project:
A basic CNN with multiple convolutional layers and max pooling to extract features.
An augmented model to enhance the dataset and improve generalization through techniques like rotation, zoom, and horizontal flipping.
Performance metrics such as accuracy, precision, and recall are computed to evaluate the models. The use of TensorBoard allows for real-time tracking of training progress and comparison of model performance metrics. Integration with Weights & Biases provides further insights through tracking experiments, saving models, and visualizing outcomes.
This project not only demonstrates the application of CNNs in image classification but also provides a framework for experimenting with different architectures and preprocessing techniques to achieve optimal results in computer vision tasks.
