Hugging Face's logo

qualcomm
/
HRNetFace

Object Detection
PyTorch
real_time
android
Model card Files
xet
 Community
HRNetFace
File size: 12,287 Bytes
ba41173
 
 
 
 
 
 
 
 
 
 
 
1529439
ba41173
 
 
1529439
c98531b
1529439
 
 
 
 
 
 
 
 
 
 
 
c98531b
 
 
 
 
 
1529439
 
 
 
 
 
c98531b
1529439
 
 
 
 
 
c98531b
1529439
 
 
 
 
 
 
 
 
 
 
 
 
 
 
c98531b
 
 
 
 
 
 
 
 
 
8ac3c5c
c98531b
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
ba41173
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
---
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).