AI Paper Atlas

Research map / Video / Video generation

Image Guides Images: Consistent Video Amodal Completion with Rectified In-Context Exemplar Guidance

Xiaoyu Kong, Ketong Ren, Dongyu She, Weiming Dong, Miao Wang

CVPR 2026 · Video › Video generation · approach: Diffusion-based human image animation

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 vanilla in-context learning with a single-frame inpainted exemplar (ablation)IC-Amodal (Flux.1-Fill) · Kubric-DynamicFVD: 278.25 -> 301.83Hurt
add calibrated dual-frame exemplar rectification (ablation)IC-Amodal · Kubric-DynamicFVD: 301.83 -> 219.37Improved
add rectified exemplar-guided attention completion (ablation)IC-Amodal · Kubric-DynamicFVD: 219.37 -> 188.81Improved
add rectified exemplar-guided attention completion (ablation)IC-Amodal · Kubric-DynamicPSNR: 24.50 -> 24.71Improved
use completed object (no scene context) as exemplar input Xex (ablation)IC-Amodal · Kubric-DynamicIoU: 86.6 -> 83.1Hurt
use only visible part as exemplar input Xex instead of original frame (ablation)IC-Amodal · Kubric-DynamicFVD: 188.81 -> 231.55Hurt
set amplification weight lambda-up to 1.5 (too small) (ablation)rectified attention · KubricqualitativeHurt
set amplification weight lambda-up to 5.0 (too large) (ablation)rectified attention · KubricqualitativeHurt
set attenuation weight lambda-down to 0.7 (too large) (ablation)rectified attention · KubricqualitativeHurt
blend raw completion with the original visible region (IC-Amodal+)IC-Amodal · Kubric-DynamicPSNR: 24.705 -> 30.091Improved
use IC-Amodal instead of fine-tuned TACO (training-free vs finetuned)IC-Amodal · Kubric-DynamicFVD: 209.28 -> 188.81Improved
use IC-Amodal instead of TACOIC-Amodal · Kubric-StaticFVD: 162.91 -> 184.10Hurt

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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