Classifier-Free Guidance
التوجيه بلا مُصنِّف
تقنية تُعزز جودة التوليد المشروط بمزج تنبؤات النموذج المشروطة وغير المشروطة أثناء الاستنتاج، دون الحاجة إلى مُصنِّف خارجي مُدرَّب.
A technique that enhances conditional generation quality by blending the model's conditional and unconditional predictions during inference, without needing a separately trained classifier.
Also translated asالإرشاد بدون مُصنِّف، التوجيه الحر من المصنِّف
First appears in this corpus in: Diffusion Models Beat GANs on Image Synthesis (2021)
Appears in these papers
- Classifier-Free Diffusion Guidance2022in the sky ✦
- Classifier-Free Diffusion Guidance2022in the sky ✦
- Consistency Models2023in the sky ✦
- Diffusion Models Beat GANs on Image Synthesis2021in the sky ✦
- Diffusion Models Beat GANs on Image Synthesis2021in the sky ✦
- GLIDE: Towards Photorealistic Image Generation and Editing with Text-Guided Diffusion Models2021in the sky ✦
- GLIDE: Towards Photorealistic Image Generation and Editing with Text-Guided Diffusion Models2021in the sky ✦
- Imagen: Photorealistic Text-to-Image Diffusion Models with Deep Language Understanding2022in the sky ✦
- Imagen: Photorealistic Text-to-Image Diffusion Models with Deep Language Understanding2022in the sky ✦
- High-Resolution Image Synthesis with Latent Diffusion Models2022in the sky ✦
- Video Generation Models as World Simulators2024in the sky ✦