AI 용어집
인공지능 완전 사전
Convolutional Neural Network (CNN)
Deep learning architecture specialized in processing grid-like data such as images, using convolutional layers, pooling layers, and fully connected layers to learn feature hierarchies.
Filter (or Kernel)
Small matrix of learnable weights that slides across the input of a convolutional layer to detect specific patterns such as edges, textures, or shapes.
Receptive Field
Region of the input image that influences a particular neuron in a given layer, increasing in size as we progress through the deeper layers of the network.
Object Detection
Task that involves locating and identifying one or more objects in an image, typically by drawing bounding boxes around them and associating each with a class.
Encoder-Decoder Architecture
Network structure composed of an encoding path that reduces spatial dimensionality to capture context, and a decoding path that increases it to produce an output of the same size as the input, used for segmentation.
Capsule Neural Network (CapsNet)
Alternative to CNNs that organizes neurons into 'capsules' to preserve the spatial hierarchy of features, aiming to better handle pose variations and object rotations.
Geometric Computer Vision
Branch of computer vision that focuses on inferring 3D properties and structures from 2D images, including tasks such as 3D reconstruction and camera pose estimation.
Vision Transformer (ViT)
Architecture that applies the transformer attention mechanism, originally designed for NLP, to sequences of image patches, competing with CNNs on classification tasks.