---
library_name: pytorch
license: mit
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/v0.64.0/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 2.50, ONNX Runtime 1.30.0 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/sixd_repnet/releases/v0.64.0/sixd_repnet-onnx-float.zip)
| QNN_DLC | float | Universal | QAIRT 2.50 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/sixd_repnet/releases/v0.64.0/sixd_repnet-qnn_dlc-float.zip)
| TFLITE | float | Universal | QAIRT 2.50 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/sixd_repnet/releases/v0.64.0/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/v0.64.0/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/v0.64.0/src/qai_hub_models/models/sixd_repnet) for usage instructions.

## Model Details

**Model Type:** Model_use_case.pose_estimation

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

## Performance Summary
| Model | Runtime | Precision | Chipset | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit
|---|---|---|---|---|---|---
| face_detector | ONNX | float | Snapdragon® 8 Elite Gen 5 For Galaxy Mobile | 1.882 ms | 2 - 149 MB | NPU
| face_detector | ONNX | float | Snapdragon® 8 Elite For Galaxy Mobile | 2.062 ms | 0 - 147 MB | NPU
| face_detector | ONNX | float | Snapdragon® X2 Elite | 1.951 ms | 5 - 5 MB | NPU
| face_detector | ONNX | float | Snapdragon® X Elite | 3.815 ms | 5 - 5 MB | NPU
| face_detector | ONNX | float | Snapdragon® 8 Gen 3 Mobile | 2.548 ms | 3 - 163 MB | NPU
| face_detector | ONNX | float | Snapdragon® 8 Gen 1 Mobile | 5.672 ms | 7 - 172 MB | NPU
| face_detector | ONNX | float | Qualcomm® Dragonwing™ IQ-8275 | 4.402 ms | 4 - 13 MB | NPU
| face_detector | ONNX | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 3.77 ms | 0 - 2 MB | NPU
| face_detector | ONNX | float | Qualcomm® QCS8450 | 5.672 ms | 7 - 172 MB | NPU
| face_detector | ONNX | float | Qualcomm® Dragonwing™ IQ-9075 | 4.531 ms | 5 - 13 MB | NPU
| face_detector | ONNX | float | Qualcomm® Dragonwing™ IQ-X7181 | 3.815 ms | 5 - 5 MB | NPU
| face_detector | ONNX | float | Qualcomm® Dragonwing™ Q-8750 | 2.062 ms | 0 - 147 MB | NPU
| face_detector | QNN_DLC | float | Snapdragon® 8 Elite Gen 5 For Galaxy Mobile | 2.241 ms | 5 - 151 MB | NPU
| face_detector | QNN_DLC | float | Snapdragon® 8 Elite For Galaxy Mobile | 2.789 ms | 0 - 144 MB | NPU
| face_detector | QNN_DLC | float | Snapdragon® X2 Elite | 2.583 ms | 5 - 5 MB | NPU
| face_detector | QNN_DLC | float | Snapdragon® X Elite | 5.438 ms | 5 - 5 MB | NPU
| face_detector | QNN_DLC | float | Snapdragon® 8 Gen 3 Mobile | 3.459 ms | 5 - 166 MB | NPU
| face_detector | QNN_DLC | float | Snapdragon® 8 Gen 1 Mobile | 8.221 ms | 3 - 63 MB | NPU
| face_detector | QNN_DLC | float | Qualcomm® Dragonwing™ IQ-8275 | 5.357 ms | 5 - 13 MB | NPU
| face_detector | QNN_DLC | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 4.942 ms | 5 - 8 MB | NPU
| face_detector | QNN_DLC | float | Qualcomm® SA8650P | 6.345 ms | 0 - 152 MB | NPU
| face_detector | QNN_DLC | float | Qualcomm® SA8255P | 6.345 ms | 0 - 152 MB | NPU
| face_detector | QNN_DLC | float | Qualcomm® QCS8450 | 8.221 ms | 3 - 63 MB | NPU
| face_detector | QNN_DLC | float | Qualcomm® Dragonwing™ IQ-9075 | 5.983 ms | 5 - 13 MB | NPU
| face_detector | QNN_DLC | float | Qualcomm® Dragonwing™ IQ-X7181 | 5.438 ms | 5 - 5 MB | NPU
| face_detector | QNN_DLC | float | Qualcomm® Dragonwing™ Q-8750 | 2.789 ms | 0 - 144 MB | NPU
| face_detector | QNN_DLC | float | Qualcomm® SA8295P | 7.513 ms | 0 - 35 MB | NPU
| face_detector | TFLITE | float | Snapdragon® 8 Elite Gen 5 For Galaxy Mobile | 2.003 ms | 0 - 146 MB | NPU
| face_detector | TFLITE | float | Snapdragon® 8 Elite For Galaxy Mobile | 2.622 ms | 0 - 148 MB | NPU
| face_detector | TFLITE | float | Snapdragon® 8 Gen 3 Mobile | 3.519 ms | 1 - 160 MB | NPU
| face_detector | TFLITE | float | Snapdragon® 8 Gen 1 Mobile | 8.249 ms | 1 - 59 MB | NPU
| face_detector | TFLITE | float | Qualcomm® Dragonwing™ IQ-8275 | 5.401 ms | 1 - 11 MB | NPU
| face_detector | TFLITE | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 4.99 ms | 1 - 3 MB | NPU
| face_detector | TFLITE | float | Qualcomm® SA8650P | 6.339 ms | 1 - 154 MB | NPU
| face_detector | TFLITE | float | Qualcomm® SA8255P | 6.339 ms | 1 - 154 MB | NPU
| face_detector | TFLITE | float | Qualcomm® QCS8450 | 8.249 ms | 1 - 59 MB | NPU
| face_detector | TFLITE | float | Qualcomm® Dragonwing™ IQ-9075 | 6.032 ms | 1 - 10 MB | NPU
| face_detector | TFLITE | float | Qualcomm® Dragonwing™ Q-8750 | 2.622 ms | 0 - 148 MB | NPU
| face_detector | TFLITE | float | Qualcomm® SA8295P | 7.562 ms | 1 - 36 MB | NPU
| pose_estimator | ONNX | float | Snapdragon® 8 Elite Gen 5 For Galaxy Mobile | 1.316 ms | 0 - 29 MB | NPU
| pose_estimator | ONNX | float | Snapdragon® 8 Elite For Galaxy Mobile | 1.442 ms | 0 - 29 MB | NPU
| pose_estimator | ONNX | float | Snapdragon® X2 Elite | 1.295 ms | 2 - 2 MB | NPU
| pose_estimator | ONNX | float | Snapdragon® X Elite | 2.452 ms | 76 - 76 MB | NPU
| pose_estimator | ONNX | float | Snapdragon® 8 Gen 3 Mobile | 1.789 ms | 0 - 41 MB | NPU
| pose_estimator | ONNX | float | Snapdragon® 8 Gen 1 Mobile | 4.314 ms | 0 - 43 MB | NPU
| pose_estimator | ONNX | float | Qualcomm® Dragonwing™ IQ-8275 | 5.026 ms | 0 - 5 MB | NPU
| pose_estimator | ONNX | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 2.431 ms | 0 - 272 MB | NPU
| pose_estimator | ONNX | float | Qualcomm® QCS8450 | 4.314 ms | 0 - 43 MB | NPU
| pose_estimator | ONNX | float | Qualcomm® Dragonwing™ IQ-9075 | 4.446 ms | 0 - 4 MB | NPU
| pose_estimator | ONNX | float | Qualcomm® Dragonwing™ IQ-X7181 | 2.452 ms | 76 - 76 MB | NPU
| pose_estimator | ONNX | float | Qualcomm® Dragonwing™ Q-8750 | 1.442 ms | 0 - 29 MB | NPU
| pose_estimator | QNN_DLC | float | Snapdragon® 8 Elite Gen 5 For Galaxy Mobile | 1.367 ms | 1 - 30 MB | NPU
| pose_estimator | QNN_DLC | float | Snapdragon® 8 Elite For Galaxy Mobile | 1.574 ms | 1 - 28 MB | NPU
| pose_estimator | QNN_DLC | float | Snapdragon® X2 Elite | 1.457 ms | 1 - 1 MB | NPU
| pose_estimator | QNN_DLC | float | Snapdragon® X Elite | 2.872 ms | 1 - 1 MB | NPU
| pose_estimator | QNN_DLC | float | Snapdragon® 8 Gen 3 Mobile | 1.95 ms | 0 - 41 MB | NPU
| pose_estimator | QNN_DLC | float | Snapdragon® 8 Gen 1 Mobile | 4.929 ms | 0 - 45 MB | NPU
| pose_estimator | QNN_DLC | float | Qualcomm® Dragonwing™ IQ-8275 | 5.123 ms | 1 - 4 MB | NPU
| pose_estimator | QNN_DLC | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 2.669 ms | 1 - 2 MB | NPU
| pose_estimator | QNN_DLC | float | Qualcomm® SA8650P | 4.812 ms | 1 - 26 MB | NPU
| pose_estimator | QNN_DLC | float | Qualcomm® SA8255P | 4.812 ms | 1 - 26 MB | NPU
| pose_estimator | QNN_DLC | float | Qualcomm® QCS8450 | 4.929 ms | 0 - 45 MB | NPU
| pose_estimator | QNN_DLC | float | Qualcomm® Dragonwing™ IQ-9075 | 4.638 ms | 3 - 5 MB | NPU
| pose_estimator | QNN_DLC | float | Qualcomm® Dragonwing™ IQ-X7181 | 2.872 ms | 1 - 1 MB | NPU
| pose_estimator | QNN_DLC | float | Qualcomm® Dragonwing™ Q-8750 | 1.574 ms | 1 - 28 MB | NPU
| pose_estimator | QNN_DLC | float | Qualcomm® SA8295P | 5.206 ms | 1 - 26 MB | NPU
| pose_estimator | TFLITE | float | Snapdragon® 8 Elite Gen 5 For Galaxy Mobile | 1.366 ms | 0 - 33 MB | NPU
| pose_estimator | TFLITE | float | Snapdragon® 8 Elite For Galaxy Mobile | 1.553 ms | 0 - 33 MB | NPU
| pose_estimator | TFLITE | float | Snapdragon® 8 Gen 3 Mobile | 1.945 ms | 0 - 49 MB | NPU
| pose_estimator | TFLITE | float | Snapdragon® 8 Gen 1 Mobile | 4.937 ms | 0 - 51 MB | NPU
| pose_estimator | TFLITE | float | Qualcomm® Dragonwing™ IQ-8275 | 5.134 ms | 0 - 79 MB | NPU
| pose_estimator | TFLITE | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 2.633 ms | 0 - 2 MB | NPU
| pose_estimator | TFLITE | float | Qualcomm® SA8650P | 4.812 ms | 0 - 27 MB | NPU
| pose_estimator | TFLITE | float | Qualcomm® SA8255P | 4.812 ms | 0 - 27 MB | NPU
| pose_estimator | TFLITE | float | Qualcomm® QCS8450 | 4.937 ms | 0 - 51 MB | NPU
| pose_estimator | TFLITE | float | Qualcomm® Dragonwing™ IQ-9075 | 4.635 ms | 0 - 78 MB | NPU
| pose_estimator | TFLITE | float | Qualcomm® Dragonwing™ Q-8750 | 1.553 ms | 0 - 33 MB | NPU
| pose_estimator | TFLITE | float | Qualcomm® SA8295P | 5.235 ms | 0 - 26 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).