AI 용어집
인공지능 완전 사전
Conditional GANs (cGAN)
GANs that allow control over generated features by using specific labels or conditions during training.
StyleGAN and StyleGAN2
Advanced GAN architecture enabling granular control over the stylistic characteristics of high-resolution generated images.
CycleGAN
GAN for image translation between domains without requiring paired data, using generation cycles.
DCGAN (Deep Convolutional GAN)
Fundamental architecture using deep convolutional networks to generate images with improved stability.
Wasserstein GAN (WGAN)
GAN variant using the Wasserstein distance to improve training stability and avoid mode collapse.
Progressive GAN (ProGAN)
Technique generating images progressively from low to high resolution to improve quality and stability.
GANs for Super-Resolution
Specialized application of GANs to enhance the resolution and quality of low-resolution images.
GANs for Face Synthesis
GANs specialized in generating photorealistic human faces with controllable variations.
GANs for Style Transfer
Application of GANs to apply artistic or visual styles to images while preserving their content.
GANs for Colorization
GANs specialized in realistically adding colors to black and white or sepia images.
GANs for Medical Imaging
Application of GANs for the generation and enhancement of medical images such as MRIs, CT scans, and X-rays.
GANs for Anomaly Detection
Using GANs to identify anomalies or defects by learning the distribution of normal data.
GANs for Video Generation
Extending GANs to temporal data to generate coherent and realistic video sequences.
GANs for Art and Design
Creative application of GANs to generate original artworks, graphic designs, and visual content.
Pix2Pix GAN
GAN for supervised image-to-image translation, transforming sketches into realistic images or modifying attributes.