Session on fundamentals of Convolutional Neural Network

The session drew AI and ML enthusiasts eager to explore how machines interpret images and extract meaningful patterns. Participants were introduced to the core concepts of computer vision about how images are represented and the tasks of feature extraction, detection, and classification. The session explored traditional methods of edge, corner, and texture detection, before moving on to modern approaches using neurons and neural networks, with examples of different network architectures. The workshop provided a deep dive into the architecture covering concepts of convolution, activation functions, pooling, flattening, fully connected layers, softmax, optimizers, and loss functions. Participants engaged in a hands-on implementation of a vanilla CNN on the MNIST dataset, training a model to recognize handwritten digits. This practical session allowed attendees to connect theoretical knowledge with real-world application. The Neural Hive club organized Pixels2Patterns, a hands-on workshop on fundamentals of Convolutional Neural Network (CNN) and its real-world applications was held on August 20, 2025 at RR campus.