Unverified50% confidenceFactExact time
UNet was proposed by Ronneberger et al. in 2015 for medical image segmentation tasks and is the core architecture of the Stable Diffusion 1.x/2.x series.
1
Sources
50%
Confidence
Long-term
Relevance
7/24/2026
First Seen
Sources
DiffusionGemma角色动作迁移实战：ComfyUI工作流拆解
redditr/comfyui7/12/2026
Related Claims
UnverifiedStable Diffusion XL采用基础模型加精炼模型的两阶段级联架构，先生成低分辨率底图再补充细节纹理63% similarUnverifiedSDXL（Stable Diffusion XL）是Stability AI推出的高分辨率图像生成模型，其完整管线包含Base模型和Refiner模型两个阶段57% similarUnverifiedMidjourney V6、DALL·E 3、Stable Diffusion 3和Flux等新一代模型能生成细节丰富、光照自然、构图精确的图像56% similarUnverifiedCLIP由OpenAI于2021年提出，SDXL引入双CLIP架构（OpenCLIP ViT-bigG + CLIP ViT-L），Stable Diffusion 3和Flux引入拥有110亿参数的T5-XXL54% similarUnverifiedDiffusion Transformer (DiT) is the dominant architecture in the current image generation field, used in models such as Stable Diffusion 3 and FLUX.54% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/608491API
curl https://kongchang.com/api/v1/knowledge/claims/608491MCP
get_claim(id=608491)