---
library_name: pytorch
license: other
tags:
- real_time
- android
pipeline_tag: keypoint-detection

---

![](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/sixd_repnet/web-assets/model_demo.png)

# SixDRepNet: Optimized for Qualcomm Devices

6DRepNet predicts head pose (pitch, yaw, roll) from a face image using a RepVGG-B1g2 backbone and a continuous 6D rotation representation, achieving robust and accurate head pose estimation.

This is based on the implementation of SixDRepNet found [here](https://github.com/thohemp/6DRepNet).
This repository contains pre-exported model files optimized for Qualcomm® devices. You can use the [Qualcomm® AI Hub Models](https://github.com/qualcomm/ai-hub-models/blob/v0.59.0/src/qai_hub_models/models/sixd_repnet) library to export with custom configurations. More details on model performance across various devices, can be found [here](#performance-summary).

Qualcomm AI Hub Models uses [Qualcomm AI Hub Workbench](https://workbench.aihub.qualcomm.com) to compile, profile, and evaluate this model. [Sign up](https://myaccount.qualcomm.com/signup) to run these models on a hosted Qualcomm® device.

## Getting Started
There are two ways to deploy this model on your device:

### Option 1: Download Pre-Exported Models

Below are pre-exported model assets ready for deployment.

| Runtime | Precision | Chipset | SDK Versions | Download |
|---|---|---|---|---|
| ONNX | float | Universal | QAIRT 2.45, ONNX Runtime 1.27.1 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/sixd_repnet/releases/v0.59.0/sixd_repnet-onnx-float.zip)
| QNN_DLC | float | Universal | QAIRT 2.45 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/sixd_repnet/releases/v0.59.0/sixd_repnet-qnn_dlc-float.zip)
| TFLITE | float | Universal | QAIRT 2.45 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/sixd_repnet/releases/v0.59.0/sixd_repnet-tflite-float.zip)

For more device-specific assets and performance metrics, visit **[SixDRepNet on Qualcomm® AI Hub](https://aihub.qualcomm.com/models/sixd_repnet)**.


### Option 2: Export with Custom Configurations

Use the [Qualcomm® AI Hub Models](https://github.com/qualcomm/ai-hub-models/blob/v0.59.0/src/qai_hub_models/models/sixd_repnet) Python library to compile and export the model with your own:
- Custom weights (e.g., fine-tuned checkpoints)
- Custom input shapes
- Target device and runtime configurations

This option is ideal if you need to customize the model beyond the default configuration provided here.

See our repository for [SixDRepNet on GitHub](https://github.com/qualcomm/ai-hub-models/blob/v0.59.0/src/qai_hub_models/models/sixd_repnet) for usage instructions.

## Model Details

**Model Type:** Model_use_case.pose_estimation

**Model Stats:**
- Input resolution: 224x224
- Number of parameters: 15.3M
- Model size (float): 58.4 MB

## Performance Summary
| Model | Runtime | Precision | Chipset | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit
|---|---|---|---|---|---|---
| face_detector | ONNX | float | Snapdragon® X2 Elite | 1.91 ms | 5 - 5 MB | NPU
| face_detector | ONNX | float | Snapdragon® X Elite | 3.66 ms | 5 - 5 MB | NPU
| face_detector | ONNX | float | Snapdragon® 8 Gen 3 Mobile | 2.163 ms | 0 - 174 MB | NPU
| face_detector | ONNX | float | Snapdragon® 8 Gen 1 Mobile | 6.801 ms | 7 - 176 MB | NPU
| face_detector | ONNX | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 3.504 ms | 0 - 5 MB | NPU
| face_detector | ONNX | float | Qualcomm® QCS8450 | 6.801 ms | 7 - 176 MB | NPU
| face_detector | ONNX | float | Qualcomm® Dragonwing™ IQ-9075 | 4.087 ms | 5 - 12 MB | NPU
| face_detector | ONNX | float | Qualcomm® Dragonwing™ IQ-X7181 | 3.66 ms | 5 - 5 MB | NPU
| face_detector | ONNX | float | Qualcomm® Dragonwing™ Q-8750 | 1.734 ms | 0 - 164 MB | NPU
| face_detector | ONNX | float | Snapdragon® 8 Elite Mobile | 1.734 ms | 0 - 164 MB | NPU
| face_detector | ONNX | float | Snapdragon® 8 Elite Gen 5 Mobile | 1.804 ms | 1 - 166 MB | NPU
| face_detector | QNN_DLC | float | Snapdragon® X2 Elite | 5.846 ms | 5 - 5 MB | NPU
| face_detector | QNN_DLC | float | Snapdragon® X Elite | 16.748 ms | 5 - 5 MB | NPU
| face_detector | QNN_DLC | float | Snapdragon® 8 Gen 3 Mobile | 9.379 ms | 5 - 174 MB | NPU
| face_detector | QNN_DLC | float | Snapdragon® 8 Gen 1 Mobile | 24.412 ms | 5 - 178 MB | NPU
| face_detector | QNN_DLC | float | Qualcomm® Dragonwing™ QCS8275 | 28.202 ms | 0 - 150 MB | NPU
| face_detector | QNN_DLC | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 15.625 ms | 5 - 86 MB | NPU
| face_detector | QNN_DLC | float | Qualcomm® SA8775P | 16.528 ms | 2 - 153 MB | NPU
| face_detector | QNN_DLC | float | Qualcomm® SA8650P | 16.528 ms | 2 - 153 MB | NPU
| face_detector | QNN_DLC | float | Qualcomm® SA8255P | 16.528 ms | 2 - 153 MB | NPU
| face_detector | QNN_DLC | float | Qualcomm® QCS8450 | 24.412 ms | 5 - 178 MB | NPU
| face_detector | QNN_DLC | float | Qualcomm® Dragonwing™ IQ-9075 | 17.407 ms | 5 - 12 MB | NPU
| face_detector | QNN_DLC | float | Qualcomm® Dragonwing™ IQ-X7181 | 16.748 ms | 5 - 5 MB | NPU
| face_detector | QNN_DLC | float | Qualcomm® Dragonwing™ Q-8750 | 6.92 ms | 0 - 154 MB | NPU
| face_detector | QNN_DLC | float | Qualcomm® SA7255P | 28.202 ms | 0 - 150 MB | NPU
| face_detector | QNN_DLC | float | Qualcomm® SA8295P | 20.611 ms | 1 - 151 MB | NPU
| face_detector | QNN_DLC | float | Snapdragon® 8 Elite Mobile | 6.92 ms | 0 - 154 MB | NPU
| face_detector | QNN_DLC | float | Snapdragon® 8 Elite Gen 5 Mobile | 5.445 ms | 4 - 157 MB | NPU
| face_detector | TFLITE | float | Snapdragon® 8 Gen 3 Mobile | 9.25 ms | 2 - 169 MB | NPU
| face_detector | TFLITE | float | Snapdragon® 8 Gen 1 Mobile | 24.502 ms | 0 - 172 MB | NPU
| face_detector | TFLITE | float | Qualcomm® Dragonwing™ QCS8275 | 28.303 ms | 1 - 151 MB | NPU
| face_detector | TFLITE | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 15.719 ms | 1 - 3 MB | NPU
| face_detector | TFLITE | float | Qualcomm® SA8775P | 16.517 ms | 1 - 154 MB | NPU
| face_detector | TFLITE | float | Qualcomm® SA8650P | 16.517 ms | 1 - 154 MB | NPU
| face_detector | TFLITE | float | Qualcomm® SA8255P | 16.517 ms | 1 - 154 MB | NPU
| face_detector | TFLITE | float | Qualcomm® QCS8450 | 24.502 ms | 0 - 172 MB | NPU
| face_detector | TFLITE | float | Qualcomm® Dragonwing™ IQ-9075 | 16.905 ms | 0 - 9 MB | NPU
| face_detector | TFLITE | float | Qualcomm® Dragonwing™ Q-8750 | 6.932 ms | 0 - 152 MB | NPU
| face_detector | TFLITE | float | Qualcomm® SA7255P | 28.303 ms | 1 - 151 MB | NPU
| face_detector | TFLITE | float | Qualcomm® SA8295P | 20.655 ms | 1 - 151 MB | NPU
| face_detector | TFLITE | float | Snapdragon® 8 Elite Mobile | 6.932 ms | 0 - 152 MB | NPU
| face_detector | TFLITE | float | Snapdragon® 8 Elite Gen 5 Mobile | 5.532 ms | 1 - 157 MB | NPU
| pose_estimator | ONNX | float | Snapdragon® X2 Elite | 1.293 ms | 2 - 2 MB | NPU
| pose_estimator | ONNX | float | Snapdragon® X Elite | 2.625 ms | 76 - 76 MB | NPU
| pose_estimator | ONNX | float | Snapdragon® 8 Gen 3 Mobile | 2.087 ms | 0 - 37 MB | NPU
| pose_estimator | ONNX | float | Snapdragon® 8 Gen 1 Mobile | 5.55 ms | 0 - 40 MB | NPU
| pose_estimator | ONNX | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 2.707 ms | 1 - 3 MB | NPU
| pose_estimator | ONNX | float | Qualcomm® QCS8450 | 5.55 ms | 0 - 40 MB | NPU
| pose_estimator | ONNX | float | Qualcomm® Dragonwing™ IQ-9075 | 4.498 ms | 0 - 4 MB | NPU
| pose_estimator | ONNX | float | Qualcomm® Dragonwing™ IQ-X7181 | 2.625 ms | 76 - 76 MB | NPU
| pose_estimator | ONNX | float | Qualcomm® Dragonwing™ Q-8750 | 1.662 ms | 0 - 24 MB | NPU
| pose_estimator | ONNX | float | Snapdragon® 8 Elite Mobile | 1.662 ms | 0 - 24 MB | NPU
| pose_estimator | ONNX | float | Snapdragon® 8 Elite Gen 5 Mobile | 1.337 ms | 0 - 26 MB | NPU
| pose_estimator | QNN_DLC | float | Snapdragon® X2 Elite | 1.527 ms | 1 - 1 MB | NPU
| pose_estimator | QNN_DLC | float | Snapdragon® X Elite | 2.869 ms | 1 - 1 MB | NPU
| pose_estimator | QNN_DLC | float | Snapdragon® 8 Gen 3 Mobile | 2.236 ms | 0 - 39 MB | NPU
| pose_estimator | QNN_DLC | float | Snapdragon® 8 Gen 1 Mobile | 6.536 ms | 0 - 40 MB | NPU
| pose_estimator | QNN_DLC | float | Qualcomm® Dragonwing™ QCS8275 | 17.831 ms | 1 - 25 MB | NPU
| pose_estimator | QNN_DLC | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 2.775 ms | 1 - 173 MB | NPU
| pose_estimator | QNN_DLC | float | Qualcomm® SA8775P | 4.828 ms | 1 - 26 MB | NPU
| pose_estimator | QNN_DLC | float | Qualcomm® SA8650P | 4.828 ms | 1 - 26 MB | NPU
| pose_estimator | QNN_DLC | float | Qualcomm® SA8255P | 4.828 ms | 1 - 26 MB | NPU
| pose_estimator | QNN_DLC | float | Qualcomm® QCS8450 | 6.536 ms | 0 - 40 MB | NPU
| pose_estimator | QNN_DLC | float | Qualcomm® Dragonwing™ IQ-9075 | 4.647 ms | 1 - 3 MB | NPU
| pose_estimator | QNN_DLC | float | Qualcomm® Dragonwing™ IQ-X7181 | 2.869 ms | 1 - 1 MB | NPU
| pose_estimator | QNN_DLC | float | Qualcomm® Dragonwing™ Q-8750 | 1.684 ms | 1 - 27 MB | NPU
| pose_estimator | QNN_DLC | float | Qualcomm® SA7255P | 17.831 ms | 1 - 25 MB | NPU
| pose_estimator | QNN_DLC | float | Qualcomm® SA8295P | 5.393 ms | 1 - 27 MB | NPU
| pose_estimator | QNN_DLC | float | Snapdragon® 8 Elite Mobile | 1.684 ms | 1 - 27 MB | NPU
| pose_estimator | QNN_DLC | float | Snapdragon® 8 Elite Gen 5 Mobile | 1.429 ms | 1 - 28 MB | NPU
| pose_estimator | TFLITE | float | Snapdragon® 8 Gen 3 Mobile | 2.244 ms | 0 - 44 MB | NPU
| pose_estimator | TFLITE | float | Snapdragon® 8 Gen 1 Mobile | 6.505 ms | 0 - 46 MB | NPU
| pose_estimator | TFLITE | float | Qualcomm® Dragonwing™ QCS8275 | 17.388 ms | 0 - 26 MB | NPU
| pose_estimator | TFLITE | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 2.849 ms | 0 - 3 MB | NPU
| pose_estimator | TFLITE | float | Qualcomm® SA8775P | 4.831 ms | 0 - 28 MB | NPU
| pose_estimator | TFLITE | float | Qualcomm® SA8650P | 4.831 ms | 0 - 28 MB | NPU
| pose_estimator | TFLITE | float | Qualcomm® SA8255P | 4.831 ms | 0 - 28 MB | NPU
| pose_estimator | TFLITE | float | Qualcomm® QCS8450 | 6.505 ms | 0 - 46 MB | NPU
| pose_estimator | TFLITE | float | Qualcomm® Dragonwing™ IQ-9075 | 4.649 ms | 0 - 78 MB | NPU
| pose_estimator | TFLITE | float | Qualcomm® Dragonwing™ Q-8750 | 1.707 ms | 0 - 31 MB | NPU
| pose_estimator | TFLITE | float | Qualcomm® SA7255P | 17.388 ms | 0 - 26 MB | NPU
| pose_estimator | TFLITE | float | Qualcomm® SA8295P | 5.349 ms | 0 - 22 MB | NPU
| pose_estimator | TFLITE | float | Snapdragon® 8 Elite Mobile | 1.707 ms | 0 - 31 MB | NPU
| pose_estimator | TFLITE | float | Snapdragon® 8 Elite Gen 5 Mobile | 1.378 ms | 0 - 27 MB | NPU

## License
* The license for the original implementation of SixDRepNet can be found
  [here](https://github.com/thohemp/6DRepNet/blob/master/LICENSE).

## References
* [6D Rotation Representation for Unconstrained Head Pose Estimation](https://arxiv.org/abs/2109.10948)
* [Source Model Implementation](https://github.com/thohemp/6DRepNet)

## Community
* Join [our AI Hub Slack community](https://aihub.qualcomm.com/community/slack) to collaborate, post questions and learn more about on-device AI.
* For questions or feedback please [reach out to us](mailto:ai-hub-support@qti.qualcomm.com).