AI Paper Atlas

Research map / 3D vision / 3D generation

StitchEdit: Stitching Depth Priors into Image Editors via Per-Layer Gradient Probing

Qin Guo, Dan Xu

NeurIPS 2026 · 3D vision › 3D generation · approach: Score distillation for text-to-3D · 0 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Proposes StitchEdit, which transfers depth priors from a pretrained diffusion backbone into an image editor to improve spatially grounded text-guided edits, selecting layers via per-layer gradient probing.

Problem

Text-guided image editing with diffusion models struggles with spatial edits (object manipulation, viewpoint changes, occlusion handling) that require 3D scene structure understanding.

Approach

Uses per-layer gradient probing to identify which layers of a diffusion backbone carry geometric (depth) priors, and stitches those depth-relevant layers into an image editor to redirect Marigold-style priors toward editing.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
fine-tune depth adapter with probe-guided per-layer schedule instead of uniformly (ablation)StitchEdit · StitchEditBenchDCS: 0.80 -> 0.92Improved
add Stage 1 depth initialization (ablation)editing LoRA · StitchEditBenchDCS: 0.70 -> 0.80Improved
replace hard top-K cutoff with probe-guided schedule (ablation)StitchEdit · StitchEditBenchDCS: 0.87 -> 0.92Improved
use probe-derived layer ordering instead of random schedule (ablation)StitchEdit · StitchEditBenchDCS: 0.83 -> 0.92Improved
use probe-computed boundaries instead of fixed cosine heuristic (ablation)StitchEdit · StitchEditBenchDCS: 0.90 -> 0.92Improved
use cosine decay instead of step or linear (ablation)StitchEdit · StitchEditBenchDCS: 0.88 / 0.90 -> 0.92Improved
train conflict-zone layers (l_s=30, l_e=55) (ablation)StitchEdit · StitchEditBenchCLIP-Dir: 0.81 -> 0.75Hurt
narrow the compatible region (l_s=10) (ablation)StitchEdit · StitchEditBenchDCS: 0.92 -> 0.86Hurt
condition on explicit source depth via ControlNet (ablation)ControlNet-Depth · StitchEditBenchDCS / CLIP-Dir: 0.72 DCS, 0.62 CLIP-DirHurt
use oracle target depth ControlNet instead of implicit prior (ablation)ControlNet-Depth · StitchEditBenchDCS: 0.92 -> 0.90Hurt
use StitchEdit instead of instruction-based editorsStitchEdit · StitchEditBenchDCS: 0.74 -> 0.92 vs Nano-BananaImproved
add Stage 1 depth LoRA to backboneQwen-Image-Edit · NYU Depth V2AbsRel: 12.2 -> 5.4Improved

Experimental setup

Builds on: Marigold

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