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

---

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

# HRNetFace: Optimized for Qualcomm Devices

Detects attributes (liveness, eye closeness, mask presence, glasses presence, sunglasses presence) that apply to a given face.

This is based on the implementation of HRNetFace found [here](https://github.com/HRNet/HRNet-Facial-Landmark-Detection).
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/hrnet_face) 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/hrnet_face/releases/v0.59.0/hrnet_face-onnx-float.zip)
| ONNX | w8a8 | Universal | QAIRT 2.45, ONNX Runtime 1.27.1 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/hrnet_face/releases/v0.59.0/hrnet_face-onnx-w8a8.zip)
| QNN_DLC | float | Universal | QAIRT 2.45 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/hrnet_face/releases/v0.59.0/hrnet_face-qnn_dlc-float.zip)
| QNN_DLC | w8a8 | Universal | QAIRT 2.45 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/hrnet_face/releases/v0.59.0/hrnet_face-qnn_dlc-w8a8.zip)
| TFLITE | float | Universal | QAIRT 2.45 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/hrnet_face/releases/v0.59.0/hrnet_face-tflite-float.zip)
| TFLITE | w8a8 | Universal | QAIRT 2.45 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/hrnet_face/releases/v0.59.0/hrnet_face-tflite-w8a8.zip)

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


### 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/hrnet_face) 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 [HRNetFace on GitHub](https://github.com/qualcomm/ai-hub-models/blob/v0.59.0/src/qai_hub_models/models/hrnet_face) for usage instructions.

## Model Details

**Model Type:** Model_use_case.object_detection

**Model Stats:**
- Model checkpoint: HR18-COFW.pth
- Input resolution: 256x256
- Number of parameters: 9.68M
- Model size (float): 36.87MB
- Model size (w8a8): 17.7 MB

## Performance Summary
| Model | Runtime | Precision | Chipset | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit
|---|---|---|---|---|---|---
| HRNetFace | ONNX | float | Snapdragon® X2 Elite | 1.551 ms | 2 - 2 MB | NPU
| HRNetFace | ONNX | float | Snapdragon® X Elite | 3.139 ms | 30 - 30 MB | NPU
| HRNetFace | ONNX | float | Snapdragon® 8 Gen 3 Mobile | 2.15 ms | 0 - 147 MB | NPU
| HRNetFace | ONNX | float | Snapdragon® 8 Gen 1 Mobile | 4.539 ms | 1 - 132 MB | NPU
| HRNetFace | ONNX | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 2.987 ms | 0 - 34 MB | NPU
| HRNetFace | ONNX | float | Qualcomm® QCS8450 | 4.539 ms | 1 - 132 MB | NPU
| HRNetFace | ONNX | float | Qualcomm® Dragonwing™ IQ-9075 | 4.997 ms | 2 - 4 MB | NPU
| HRNetFace | ONNX | float | Qualcomm® Dragonwing™ IQ-X7181 | 3.139 ms | 30 - 30 MB | NPU
| HRNetFace | ONNX | float | Qualcomm® Dragonwing™ Q-8750 | 1.65 ms | 0 - 92 MB | NPU
| HRNetFace | ONNX | float | Snapdragon® 8 Elite Mobile | 1.65 ms | 0 - 92 MB | NPU
| HRNetFace | ONNX | float | Snapdragon® 8 Elite Gen 5 Mobile | 1.352 ms | 0 - 92 MB | NPU
| HRNetFace | ONNX | w8a8 | Snapdragon® X2 Elite | 0.663 ms | 1 - 1 MB | NPU
| HRNetFace | ONNX | w8a8 | Snapdragon® X Elite | 1.331 ms | 16 - 16 MB | NPU
| HRNetFace | ONNX | w8a8 | Snapdragon® 8 Gen 3 Mobile | 0.875 ms | 0 - 156 MB | NPU
| HRNetFace | ONNX | w8a8 | Snapdragon® 8 Gen 1 Mobile | 1.733 ms | 0 - 158 MB | NPU
| HRNetFace | ONNX | w8a8 | Qualcomm® Dragonwing™ QCS6490 | 4.98 ms | 0 - 3 MB | NPU
| HRNetFace | ONNX | w8a8 | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 1.289 ms | 0 - 18 MB | NPU
| HRNetFace | ONNX | w8a8 | Qualcomm® QCS8450 | 1.733 ms | 0 - 158 MB | NPU
| HRNetFace | ONNX | w8a8 | Qualcomm® Dragonwing™ IQ-9075 | 1.498 ms | 0 - 3 MB | NPU
| HRNetFace | ONNX | w8a8 | Qualcomm® Dragonwing™ IQ-X7181 | 1.331 ms | 16 - 16 MB | NPU
| HRNetFace | ONNX | w8a8 | Qualcomm® Dragonwing™ Q-8750 | 0.705 ms | 0 - 109 MB | NPU
| HRNetFace | ONNX | w8a8 | Snapdragon® 8 Elite Mobile | 0.705 ms | 0 - 109 MB | NPU
| HRNetFace | ONNX | w8a8 | Snapdragon® 8 Elite Gen 5 Mobile | 0.593 ms | 0 - 113 MB | NPU
| HRNetFace | QNN_DLC | float | Snapdragon® X2 Elite | 1.963 ms | 1 - 1 MB | NPU
| HRNetFace | QNN_DLC | float | Snapdragon® X Elite | 3.692 ms | 1 - 1 MB | NPU
| HRNetFace | QNN_DLC | float | Snapdragon® 8 Gen 3 Mobile | 2.333 ms | 0 - 120 MB | NPU
| HRNetFace | QNN_DLC | float | Snapdragon® 8 Gen 1 Mobile | 5.487 ms | 0 - 105 MB | NPU
| HRNetFace | QNN_DLC | float | Qualcomm® Dragonwing™ QCS8275 | 15.432 ms | 1 - 75 MB | NPU
| HRNetFace | QNN_DLC | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 3.288 ms | 0 - 10 MB | NPU
| HRNetFace | QNN_DLC | float | Qualcomm® SA8775P | 5.119 ms | 1 - 79 MB | NPU
| HRNetFace | QNN_DLC | float | Qualcomm® SA8650P | 5.119 ms | 1 - 79 MB | NPU
| HRNetFace | QNN_DLC | float | Qualcomm® SA8255P | 5.119 ms | 1 - 79 MB | NPU
| HRNetFace | QNN_DLC | float | Qualcomm® QCS8450 | 5.487 ms | 0 - 105 MB | NPU
| HRNetFace | QNN_DLC | float | Qualcomm® Dragonwing™ IQ-9075 | 4.827 ms | 1 - 4 MB | NPU
| HRNetFace | QNN_DLC | float | Qualcomm® Dragonwing™ IQ-X7181 | 3.692 ms | 1 - 1 MB | NPU
| HRNetFace | QNN_DLC | float | Qualcomm® Dragonwing™ Q-8750 | 1.754 ms | 0 - 78 MB | NPU
| HRNetFace | QNN_DLC | float | Qualcomm® SA7255P | 15.432 ms | 1 - 75 MB | NPU
| HRNetFace | QNN_DLC | float | Qualcomm® SA8295P | 5.641 ms | 1 - 62 MB | NPU
| HRNetFace | QNN_DLC | float | Snapdragon® 8 Elite Mobile | 1.754 ms | 0 - 78 MB | NPU
| HRNetFace | QNN_DLC | float | Snapdragon® 8 Elite Gen 5 Mobile | 1.411 ms | 1 - 82 MB | NPU
| HRNetFace | QNN_DLC | w8a8 | Snapdragon® X2 Elite | 0.839 ms | 0 - 0 MB | NPU
| HRNetFace | QNN_DLC | w8a8 | Snapdragon® X Elite | 1.552 ms | 0 - 0 MB | NPU
| HRNetFace | QNN_DLC | w8a8 | Snapdragon® 8 Gen 3 Mobile | 0.906 ms | 0 - 109 MB | NPU
| HRNetFace | QNN_DLC | w8a8 | Snapdragon® 8 Gen 1 Mobile | 1.844 ms | 0 - 108 MB | NPU
| HRNetFace | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ QCS6490 | 5.726 ms | 2 - 4 MB | NPU
| HRNetFace | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ QCS8275 | 3.235 ms | 0 - 73 MB | NPU
| HRNetFace | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 1.334 ms | 0 - 2 MB | NPU
| HRNetFace | QNN_DLC | w8a8 | Qualcomm® SA8775P | 1.696 ms | 0 - 74 MB | NPU
| HRNetFace | QNN_DLC | w8a8 | Qualcomm® SA8650P | 1.696 ms | 0 - 74 MB | NPU
| HRNetFace | QNN_DLC | w8a8 | Qualcomm® SA8255P | 1.696 ms | 0 - 74 MB | NPU
| HRNetFace | QNN_DLC | w8a8 | Qualcomm® QCS8450 | 1.844 ms | 0 - 108 MB | NPU
| HRNetFace | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ IQ-9075 | 1.515 ms | 0 - 2 MB | NPU
| HRNetFace | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ IQ-X7181 | 1.552 ms | 0 - 0 MB | NPU
| HRNetFace | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ Q-6690 | 10.603 ms | 0 - 199 MB | NPU
| HRNetFace | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ Q-7790 | 1.622 ms | 0 - 79 MB | NPU
| HRNetFace | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ Q-8750 | 0.678 ms | 0 - 69 MB | NPU
| HRNetFace | QNN_DLC | w8a8 | Qualcomm® SA7255P | 3.235 ms | 0 - 73 MB | NPU
| HRNetFace | QNN_DLC | w8a8 | Qualcomm® SA8295P | 2.199 ms | 0 - 70 MB | NPU
| HRNetFace | QNN_DLC | w8a8 | Snapdragon® 8 Elite Mobile | 0.678 ms | 0 - 69 MB | NPU
| HRNetFace | QNN_DLC | w8a8 | Snapdragon® 8 Elite Gen 5 Mobile | 0.53 ms | 0 - 73 MB | NPU
| HRNetFace | QNN_DLC | w8a8 | Snapdragon® 7 Gen 4 Mobile | 1.622 ms | 0 - 79 MB | NPU
| HRNetFace | TFLITE | float | Snapdragon® 8 Gen 3 Mobile | 2.345 ms | 0 - 139 MB | NPU
| HRNetFace | TFLITE | float | Snapdragon® 8 Gen 1 Mobile | 5.464 ms | 0 - 126 MB | NPU
| HRNetFace | TFLITE | float | Qualcomm® Dragonwing™ QCS8275 | 15.477 ms | 1 - 92 MB | NPU
| HRNetFace | TFLITE | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 3.327 ms | 0 - 3 MB | NPU
| HRNetFace | TFLITE | float | Qualcomm® SA8775P | 5.117 ms | 1 - 93 MB | NPU
| HRNetFace | TFLITE | float | Qualcomm® SA8650P | 5.117 ms | 1 - 93 MB | NPU
| HRNetFace | TFLITE | float | Qualcomm® SA8255P | 5.117 ms | 1 - 93 MB | NPU
| HRNetFace | TFLITE | float | Qualcomm® QCS8450 | 5.464 ms | 0 - 126 MB | NPU
| HRNetFace | TFLITE | float | Qualcomm® Dragonwing™ IQ-9075 | 4.86 ms | 0 - 35 MB | NPU
| HRNetFace | TFLITE | float | Qualcomm® Dragonwing™ Q-8750 | 1.759 ms | 0 - 94 MB | NPU
| HRNetFace | TFLITE | float | Qualcomm® SA7255P | 15.477 ms | 1 - 92 MB | NPU
| HRNetFace | TFLITE | float | Qualcomm® SA8295P | 5.696 ms | 1 - 74 MB | NPU
| HRNetFace | TFLITE | float | Snapdragon® 8 Elite Mobile | 1.759 ms | 0 - 94 MB | NPU
| HRNetFace | TFLITE | float | Snapdragon® 8 Elite Gen 5 Mobile | 1.408 ms | 0 - 90 MB | NPU
| HRNetFace | TFLITE | w8a8 | Snapdragon® 8 Gen 3 Mobile | 0.736 ms | 0 - 111 MB | NPU
| HRNetFace | TFLITE | w8a8 | Snapdragon® 8 Gen 1 Mobile | 1.561 ms | 0 - 112 MB | NPU
| HRNetFace | TFLITE | w8a8 | Qualcomm® Dragonwing™ QCS6490 | 5.28 ms | 0 - 18 MB | NPU
| HRNetFace | TFLITE | w8a8 | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 1.11 ms | 0 - 3 MB | NPU
| HRNetFace | TFLITE | w8a8 | Qualcomm® SA8775P | 1.512 ms | 0 - 74 MB | NPU
| HRNetFace | TFLITE | w8a8 | Qualcomm® SA8650P | 1.512 ms | 0 - 74 MB | NPU
| HRNetFace | TFLITE | w8a8 | Qualcomm® SA8255P | 1.512 ms | 0 - 74 MB | NPU
| HRNetFace | TFLITE | w8a8 | Qualcomm® QCS8450 | 1.561 ms | 0 - 112 MB | NPU
| HRNetFace | TFLITE | w8a8 | Qualcomm® Dragonwing™ IQ-9075 | 1.537 ms | 0 - 18 MB | NPU
| HRNetFace | TFLITE | w8a8 | Qualcomm® Dragonwing™ Q-6690 | 9.831 ms | 0 - 192 MB | NPU
| HRNetFace | TFLITE | w8a8 | Qualcomm® Dragonwing™ Q-7790 | 1.354 ms | 0 - 75 MB | NPU
| HRNetFace | TFLITE | w8a8 | Qualcomm® Dragonwing™ Q-8750 | 0.579 ms | 0 - 68 MB | NPU
| HRNetFace | TFLITE | w8a8 | Qualcomm® SA8295P | 1.99 ms | 0 - 69 MB | NPU
| HRNetFace | TFLITE | w8a8 | Snapdragon® 8 Elite Mobile | 0.579 ms | 0 - 68 MB | NPU
| HRNetFace | TFLITE | w8a8 | Snapdragon® 8 Elite Gen 5 Mobile | 0.471 ms | 0 - 73 MB | NPU
| HRNetFace | TFLITE | w8a8 | Snapdragon® 7 Gen 4 Mobile | 1.354 ms | 0 - 75 MB | NPU

## License
* The license for the original implementation of HRNetFace can be found
  [here](https://github.com/HRNet/HRNet-Facial-Landmark-Detection/blob/master/LICENCE).

## References
* [Deep High-Resolution Representation Learning for Visual Recognition](https://arxiv.org/abs/1908.07919)
* [Source Model Implementation](https://github.com/HRNet/HRNet-Facial-Landmark-Detection)

## 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).