Sapiens2: 2D Human Pose Estimation
May 15, 2026 · View on GitHub
308-keypoint top-down pose estimation including detailed face (274 keypoints), hand, and foot keypoints. Predictions follow the Sociopticon keypoint format.
Model Zoo
Download checkpoints from HuggingFace
and place them under $SAPIENS_CHECKPOINT_ROOT (default: ~/sapiens2_host).
| Model | Checkpoint Path |
|---|---|
| Sapiens2-0.4B | $SAPIENS_CHECKPOINT_ROOT/pose/sapiens2_0.4b_pose.safetensors |
| Sapiens2-0.8B | $SAPIENS_CHECKPOINT_ROOT/pose/sapiens2_0.8b_pose.safetensors |
| Sapiens2-1B | $SAPIENS_CHECKPOINT_ROOT/pose/sapiens2_1b_pose.safetensors |
| Sapiens2-5B | $SAPIENS_CHECKPOINT_ROOT/pose/sapiens2_5b_pose.safetensors |
Person Detector
hf download facebook/detr-resnet-101-dc5 \
--local-dir "${SAPIENS_CHECKPOINT_ROOT}/detector/detr-resnet-101-dc5"
Inference Guide
Runs on demo set (demo/data, 100 frames) by default:
cd $SAPIENS_ROOT/sapiens/pose
./scripts/demo/keypoints308.sh
Open the script and adjust:
INPUT— path to your image directory (default:../../demo/data)OUTPUT— where to save visualizationsMODEL_NAME— uncomment the model size you want to useLINE_THICKNESS,RADIUS,KPT_THRES— visualization styleJOBS_PER_GPU,GPU_IDS— parallelism (defaults: 2 jobs/GPU on GPUs 0–7)
Outputs (visualization images + per-image keypoint JSONs) are written to $OUTPUT.
Resources
- Demo: facebook/sapiens2-pose
- Models: 0.4B, 0.8B, 1B, 5B
- Detector: facebook/detr-resnet-101-dc5