EasonXiao-888

EasonXiao-888 / SpatialEdit

Public

[Official Repo] SpatialEdit: Benchmarking Fine-Grained Image Spatial Editing

83
1
100% credibility
Found Apr 10, 2026 at 83 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
Python
AI Summary

SpatialEdit provides a benchmark, large synthetic dataset, and strong baseline model for evaluating and performing fine-grained spatial image editing tasks like object movement, rotation, and camera control.

How It Works

1
🔍 Discover SpatialEdit

You find this cool tool on Hugging Face that lets you precisely move objects or change viewpoints in photos, like rotating a dog to see its front view.

2
📥 Grab the ready examples

Download the sample images and demo files from the links provided, everything you need to start playing.

3
⚡ Try the quick demo

Run the simple demo script on your computer to instantly see a photo transformed, like making an object spin or shift position.

4
🖼️ Upload your photo

Pick any picture from your gallery and describe the change you want, such as 'move the car left' or 'zoom in on the face'.

5
✨ Watch the magic

Hit go and see your image edited exactly as described, with perfect spatial control and no unwanted changes.

6
📊 Test and compare

Use the built-in checker to score how well edits match real changes, like camera angles or object positions.

🎉 Perfect edits every time

You now create stunning, precise photo edits effortlessly, ready to share or use in your projects.

Sign up to see the full architecture

5 more

Sign Up Free

Star Growth

See how this repo grew from 83 to 83 stars Sign Up Free
Repurpose This Repo

Repurpose is a Pro feature

Generate ready-to-use prompts for X threads, LinkedIn posts, blog posts, YouTube scripts, and more -- with full repo context baked in.

Unlock Repurpose
AI-Generated Review

What is SpatialEdit?

SpatialEdit is a Python benchmark suite for evaluating fine-grained image spatial editing, tackling the gap in assessing precise camera viewpoint changes and object manipulations like moving or rotating elements. Developers get ready-to-run inference scripts, a 500K synthetic dataset for training, a 16B-parameter baseline model, and evaluation pipelines measuring perceptual quality alongside geometric accuracy via viewpoint reconstruction and framing analysis. All resources link to official Hugging Face pages for datasets, models, and benchmarks.

Why is it gaining traction?

It stands out by combining visual plausibility with hard geometric fidelity metrics—using tools like VGGT for viewpoints and YOLO for framing—where most editing benchmarks overlook spatial precision. The official repository provides distributed eval scripts via torchrun, quick demos for tasks like 3D point control and object rotation-to-video, and competitive baselines that beat priors on spatial tasks, drawing devs seeking reliable official benchmarking.

Who should use this?

Computer vision researchers benchmarking spatial editing models, AI engineers fine-tuning diffusion models for camera trajectories or object relocation, and teams in AR/VR prototyping precise layout edits. Ideal for those tired of ad-hoc evals without ground-truth geometry.

Verdict

Worth forking for standardized spatial benchmarking, especially with official HF releases and paper; low 1.0% credibility score and 81 stars signal early-stage maturity—expect rough paths in configs—but solid metrics make it a practical official repository starter.

(198 words)

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.