AI Paper Atlas

Research map / Architectures and efficiency / Attention and transformers

Sapiens2

Rawal Khirodkar, He Wen, Julieta Martinez, Yuan Dong, Zhaoen Su, Shunsuke Saito

ICLR 2026 · Architectures and efficiency › Attention and transformers · approach: Efficient and robust vision transformers · 7 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
scale Sapiens2 from 0.4B to 5B parametersSapiens2 · Pose 308 keypoints (11K test)mAP (%): 76.9 -> 82.3Improved
scale Sapiens2 from 0.4B to 5B parametersSapiens2 · Albedo estimation (10K test)MAE: 0.01825 -> 0.01191Improved
replace Sapiens-1B with Sapiens2-1B (in-the-wild supervision, 1K output)Sapiens-1B · Body-part segmentation (5K test, 29 classes)mIoU (%): 53.8 -> 81.7Improved
combine MAE with contrastive pretraining (Sapiens2-5B vs DINOv3-7B)DINOv3-7B · Dense probing (frozen backbone)pose mAP: 68.2 -> 74.7Improved
combine MAE with contrastive pretraining (Sapiens2-5B vs DINOv3-7B)DINOv3-7B · Dense probing (frozen backbone)albedo MAE (x1e-2): 3.48 -> 3.12Improved
use 4K hierarchical windowed backbone instead of 1K for 1B modelSapiens2-1B · Body-part segmentation (5K test, 29 classes)mIoU (%): 81.7 -> 81.9Improved
use 4K hierarchical windowed backbone instead of 1K for 1B modelSapiens2-1B · Normal estimation (10K whole-body test)mean angular error (deg): 7.12 -> 6.98Improved
use Sapiens2-5B instead of Sapiens2-1BSapiens2 · Body-part segmentation (5K test, 29 classes)mAcc (%): 92.0 -> 91.1Hurt
use Sapiens2-1B instead of Sapiens2-0.8BSapiens2 · Pointmap (10K test)RMSE: 0.459 -> 0.478Hurt

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