Unverified50% confidenceFactExact time
Commonly used super-resolution models include ESRGAN and Real-ESRGAN.
1
Sources
50%
Confidence
Long-term
Relevance
9/30/2026
First Seen
Sources
Related Entities
Related Claims
Unverified图像超分辨率技术从 2014 年的 SRCNN 开始，经历 EDSR、RCAN，再到基于 GAN 的 SRGAN 和 ESRGAN 的发展69% similarUnverifiedDALL·E 3基于扩散模型（Diffusion Model）架构，相较于前代GAN在图像质量和指令遵循度上有显著提升58% similarUnverifiedVHectorLab 3D integrates Top-K SAE (Sparse Autoencoder).55% similarUnverifiedThe model architecture is a cross-attention DiT (896 width × 16 layers) using 2D RoPE, QK-norm, SwiGLU, and adaLN-single.55% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/960589API
curl https://kongchang.com/api/v1/knowledge/claims/960589MCP
get_claim(id=960589)