Sapiens2-5B-Surface

Per-pixel surface-normal estimation (3-channel unit vectors in camera frame).

This repository contains the 5B Surface Normal Estimation checkpoint, finetuned from the Sapiens2-5B pretrained backbone.

📄 Paper: arXiv:2604.21681
🌐 Project Page: rawalkhirodkar.github.io/sapiens2
💻 Code: github.com/facebookresearch/sapiens2

Model Details

Developed by: Meta
Model type: Vision Transformer
License: Sapiens2 License
Task: normal
Base model: facebook/sapiens2-pretrain-5b
Format: safetensors
File: sapiens2_5b_normal.safetensors

Quick Start

Install the Sapiens2 repo (pip install -e .), download the checkpoint, and run the demo:

# 1. Download the checkpoint to $SAPIENS_CHECKPOINT_ROOT/normal/
hf download facebook/sapiens2-normal-5b sapiens2_5b_normal.safetensors \
    --local-dir ~/sapiens2_host/normal

# 2. Run the demo (edit INPUT, OUTPUT, and MODEL_NAME inside the script)
cd $SAPIENS_ROOT/sapiens/dense
./scripts/demo/normal.sh

See the Surface Normal Estimation guide for details on inputs, outputs, and visualization options.

Model Card

Field	Value
Architecture	Sapiens2 ViT backbone + Surface Normal Estimation head
Backbone parameters	5.071 B
Backbone FLOPs	15.722 T
Embedding dim	2432
Layers	56
Attention heads	32
Inference resolution	1024 × 768 (H × W)
Patch size	16

Sapiens2-Surface Family

Model	Params	FLOPs	Embed dim	Layers	Heads
Sapiens2-0.4B	0.398 B	1.260 T	1024	24	16
Sapiens2-0.8B	0.818 B	2.592 T	1280	32	16
Sapiens2-1B	1.462 B	4.715 T	1536	40	24
Sapiens2-5B (this)	5.071 B	15.722 T	2432	56	32

See the Sapiens2 Collection for all variants and other downstream task checkpoints.

Intended Use

Surface Normal Estimation on human-centric imagery
Research on human-centric vision

License

Released under the Sapiens2 License.

Citation

@article{khirodkarsapiens2,
  title={Sapiens2},
  author={Khirodkar, Rawal and Wen, He and Martinez, Julieta and Dong, Yuan and Su, Zhaoen and Saito, Shunsuke},
  journal={arXiv preprint arXiv:2604.21681},
  year={2026}
}