What is the filter called that is used to extract features from images?

The term "kernel" refers to a small matrix used in convolutional operations within convolutional neural networks (CNNs) to extract features from images. Convolutional layers apply these kernels to the input image to produce feature maps, which highlight various patterns, textures, and shapes detected in the image. The size and values of the kernel determine what specific features are being extracted. Different kernels can be designed to detect edges, corners, or other visual elements, making them essential for the performance of CNNs in tasks such as image classification, object detection, and more.

In contrast, the other terms relate to different aspects of neural networks. "Layer" refers to a collection of nodes or neurons in which computations occur, while "node" typically represents an individual neuron within a layer. A "feature map," on the other hand, is the result of applying a kernel to an input image, representing the extracted features at a certain level of abstraction. While feature maps are crucial in understanding the output, it is the kernel that initiates this feature extraction process.