Convolutional Neural Network (CNN)
A neural network architecture specialized for processing grid-like data such as images, using learned filters that detect features like edges, textures, and objects.
How CNNs Work
Convolutional layers slide small filters across the input, detecting local patterns. Pooling layers reduce spatial dimensions. Deeper layers detect increasingly complex features: edges → textures → parts → objects. Fully connected layers at the end make predictions.
Key Architectures
LeNet (1998): Pioneered CNNs for digit recognition. AlexNet (2012): Sparked the deep learning revolution. ResNet (2015): Introduced skip connections for very deep networks. EfficientNet: Optimal scaling of depth, width, resolution.
Modern Usage
While Vision Transformers (ViTs) now rival CNNs on many tasks, CNNs remain popular for edge devices (efficient computation), real-time applications, and as feature extractors in multimodal models.