Convolutional Neural Network (CNN) - Page 2

A convolutional neural network, or CNN, is built to recognize spatial patterns in grid-like data such as images. Small filters move across the input and detect local features, beginning with simple edges or textures and progressing toward shapes and objects in deeper layers. Because the same filter is reused across different positions, a CNN needs fewer parameters than a fully connected network processing the same image. Pooling or strided operations can reduce dimensions while preserving important signals. CNNs became foundational to computer vision and remain widely used for classification, object detection, medical imaging, quality inspection, and video analysis. Their predictions can still be affected by biased datasets, unfamiliar viewpoints, adversarial changes, and visual cues that differ from what people consider meaningful.