AI Paper Atlas

Research map / 3D vision / 3D generation

Ar2Can: An Architect and an Artist Leveraging a Canvas for Multi-Human Generation

Shubhankar Borse, Phuc Pham, Farzad Farhadzadeh, Seokeon Choi, Phong Nguyen, Anh Tran, Sungrack Yun, Munawar Hayat, Fatih Porikli

CVPR 2026 · 3D vision › 3D generation · approach: Diffusion-based novel view synthesis · 1 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 Hungarian centroid face-matching reward in place of simple exact-location matching (ablation)Artist (Flux-Kontext GRPO) · MultiHuman-TestBenchMulti-ID similarity: 55.2 -> 60.3Improved
add Hungarian centroid face-matching reward in place of simple exact-location matching (ablation)Artist (Flux-Kontext GRPO) · MultiHuman-TestBenchHPS: 27.6 -> 30.9Improved
add simple exact-location face-matching reward to Kontext (ablation)Flux-Kontext · MultiHuman-TestBenchCount accuracy: 80.7 -> 75.6Hurt
add simple exact-location face-matching reward to Kontext (ablation)Flux-Kontext · MultiHuman-TestBenchMulti-ID similarity: 14.5 -> 55.2Improved
add curriculum training over person count (ablation)Artist with Hungarian reward · MultiHuman-TestBenchCount accuracy: 80.1 -> 86.9Improved
add curriculum training over person count (ablation)Artist with Hungarian reward · MultiHuman-TestBenchMulti-ID similarity: 60.3 -> 68.2Improved
add curriculum training over person count (ablation)Artist with Hungarian reward · MultiHuman-TestBenchHPS: 30.9 -> 30.8No clear effect
RL-train Artist with spatially-grounded rewards on Architect layoutsFlux-Kontext · MultiHuman-TestBenchMulti-ID similarity: 14.5 -> 68.2Improved
use Architect-A (Qwen) instead of Architect-B (Flux-Schnell)Ar2Can · MultiHuman-TestBenchCount accuracy: 86.9 -> 90.2Improved
use Architect-A instead of Architect-BAr2Can · MultiHuman-TestBenchAction-C: 82.0 -> 77.6Hurt
Ar2Can vs WithAnyone on out-of-domain-biased MultiID-TestWithAnyone · MultiID-TestMulti-ID (Ref) similarity: 50.1 -> 54.3Improved
token dropping and sharing on canvasArtist · Inference latency (A100, 3 identities)latencyImproved

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