AI Paper Atlas

Research map / Image generation / Editing and style transfer

Visual Instruction Inversion: Image Editing via Image Prompting

Thao Nguyen, Yuheng Li, Utkarsh Ojha, Yong Jae Lee

NeurIPS 2023 · Image generation › Editing and style transfer · approach: Instruction-based diffusion editing · 6 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add CLIP loss to MSE loss (caption init, random noise) (ablation)Visual Instruction Inversion · Clean-InstructPix2PixVisual CLIP similarity: 0.254 -> 0.299Improved
add CLIP loss to MSE loss (caption init, random noise) (ablation)Visual Instruction Inversion · Clean-InstructPix2PixDirectional CLIP similarity: 0.115 -> 0.126Improved
add CLIP loss to MSE loss (caption init, random noise) (ablation)Visual Instruction Inversion · Clean-InstructPix2PixImage CLIP similarity: 0.845 -> 0.823Hurt
add CLIP loss (ground-truth init, random noise) (ablation)Visual Instruction Inversion · Clean-InstructPix2PixVisual CLIP similarity: 0.247 -> 0.231Hurt
initialize instruction from caption of after image instead of ground-truth instruction (MSE+CLIP, random noise) (ablation)Visual Instruction Inversion · Clean-InstructPix2PixVisual CLIP similarity: 0.231 -> 0.299Improved
initialize from caption instead of ground-truth instruction (MSE+CLIP, random noise) (ablation)Visual Instruction Inversion · Clean-InstructPix2PixDirectional CLIP similarity: 0.131 -> 0.126Mixed
use caption directly as InstructPix2Pix instruction without training (ablation)InstructPix2Pix · Clean-InstructPix2PixDirectional CLIP similarity: 0.126 -> 0.090Hurt
reuse training noise at test time (full method) (ablation)Visual Instruction Inversion · Clean-InstructPix2PixImage CLIP similarity: 0.823 -> 0.847Improved
reuse training noise at test time (full method) (ablation)Visual Instruction Inversion · Clean-InstructPix2PixVisual CLIP similarity: 0.299 -> 0.251Hurt
reuse training noise at test time (full method) (ablation)Visual Instruction Inversion · Clean-InstructPix2PixDirectional CLIP similarity: 0.126 -> 0.113Hurt
increase training pairs from 1 to 4 (random noise) (ablation)Visual Instruction Inversion · Clean-InstructPix2PixDirectional CLIP similarity: 0.126 -> 0.148Improved
increase training pairs from 1 to 4 (random noise) (ablation)Visual Instruction Inversion · Clean-InstructPix2PixImage CLIP similarity: 0.823 -> 0.798Hurt

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy