https://static.picstashpic.com/product/cover/d098545c-4a9f-4651-9075-da078fdc7678.jpg

November 2026 (A GROUP CW) - Lecture 9: Image-to-Image Generation

US$8.00

数字商品,购入后不可退款
包含 1 个链接

Core papers / methods:

  1. Latent Diffusion Models

  2. SDEdit: Guided Image Synthesis and Editing with Stochastic Differential Equations
    Meng et al., 2021

Topics:

  • Image-to-image is not direct pixel editing

  • Adding noise to an input image

  • Denoising strength

  • How prompts modify the image

  • How reference images act as conditions

Key concept:

input image → add controlled noise → conditional denoising → edited image

Engineering connection:

  • Stable Diffusion img2img

  • Seedream image editing

  • Nano Banana / reference-to-image workflows

  • Product image editing

  • Portrait generation

  • Style transfer