Facial-Attribute-Detection: Optimized for Qualcomm Devices
Detects attributes (eye closeness, mask presence, eyeglasses presence, sunglasses presence) that apply to a given face. This model's architecture was developed by Qualcomm. The model was trained by Qualcomm on a proprietary dataset of faces, but can be used on any image.
This repository contains pre-exported model files optimized for Qualcomm® devices. You can use the Qualcomm® AI Hub Models library to export with custom configurations. More details on model performance across various devices, can be found here.
Qualcomm AI Hub Models uses Qualcomm AI Hub Workbench to compile, profile, and evaluate this model. Sign up 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 2.42, ONNX Runtime 1.24.3 | Download |
| ONNX | w8a8 | Universal | QAIRT 2.42, ONNX Runtime 1.24.3 | Download |
| QNN_DLC | float | Universal | QAIRT 2.43 | Download |
| QNN_DLC | w8a8 | Universal | QAIRT 2.43 | Download |
| TFLITE | float | Universal | QAIRT 2.43, TFLite 2.19.1 | Download |
| TFLITE | w8a8 | Universal | QAIRT 2.43, TFLite 2.19.1 | Download |
For more device-specific assets and performance metrics, visit Facial-Attribute-Detection on Qualcomm® AI Hub.
Option 2: Export with Custom Configurations
Use the Qualcomm® AI Hub Models 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 Facial-Attribute-Detection on GitHub for usage instructions.
Model Details
Model Type: Model_use_case.object_detection
Model Stats:
- Model checkpoint: multitask_FR_state_dict.pt
- Input resolution: 128x128
- Number of parameters: 12.1M
- Model size (float): 46.3 MB
- Model size (w8a8): 12.3 MB
Performance Summary
| Model | Runtime | Precision | Chipset | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit |
|---|---|---|---|---|---|---|
| Facial-Attribute-Detection | ONNX | float | Snapdragon® 8 Elite Gen 5 Mobile | 0.435 ms | 0 - 65 MB | NPU |
| Facial-Attribute-Detection | ONNX | float | Snapdragon® X2 Elite | 0.415 ms | 22 - 22 MB | NPU |
| Facial-Attribute-Detection | ONNX | float | Snapdragon® X Elite | 1.004 ms | 22 - 22 MB | NPU |
| Facial-Attribute-Detection | ONNX | float | Snapdragon® 8 Gen 3 Mobile | 0.642 ms | 0 - 82 MB | NPU |
| Facial-Attribute-Detection | ONNX | float | Qualcomm® QCS8550 (Proxy) | 0.862 ms | 0 - 7 MB | NPU |
| Facial-Attribute-Detection | ONNX | float | Qualcomm® QCS9075 | 1.263 ms | 0 - 3 MB | NPU |
| Facial-Attribute-Detection | ONNX | float | Snapdragon® 8 Elite For Galaxy Mobile | 0.518 ms | 0 - 64 MB | NPU |
| Facial-Attribute-Detection | ONNX | w8a8 | Snapdragon® 8 Elite Gen 5 Mobile | 0.219 ms | 0 - 61 MB | NPU |
| Facial-Attribute-Detection | ONNX | w8a8 | Snapdragon® X2 Elite | 0.187 ms | 11 - 11 MB | NPU |
| Facial-Attribute-Detection | ONNX | w8a8 | Snapdragon® X Elite | 0.545 ms | 11 - 11 MB | NPU |
| Facial-Attribute-Detection | ONNX | w8a8 | Snapdragon® 8 Gen 3 Mobile | 0.298 ms | 0 - 86 MB | NPU |
| Facial-Attribute-Detection | ONNX | w8a8 | Qualcomm® QCS6490 | 35.379 ms | 20 - 27 MB | CPU |
| Facial-Attribute-Detection | ONNX | w8a8 | Qualcomm® QCS8550 (Proxy) | 0.398 ms | 0 - 15 MB | NPU |
| Facial-Attribute-Detection | ONNX | w8a8 | Qualcomm® QCS9075 | 0.529 ms | 0 - 3 MB | NPU |
| Facial-Attribute-Detection | ONNX | w8a8 | Qualcomm® QCM6690 | 17.115 ms | 18 - 27 MB | CPU |
| Facial-Attribute-Detection | ONNX | w8a8 | Snapdragon® 8 Elite For Galaxy Mobile | 0.247 ms | 0 - 62 MB | NPU |
| Facial-Attribute-Detection | ONNX | w8a8 | Snapdragon® 7 Gen 4 Mobile | 12.593 ms | 19 - 29 MB | CPU |
| Facial-Attribute-Detection | QNN_DLC | float | Snapdragon® 8 Elite Gen 5 Mobile | 0.412 ms | 0 - 58 MB | NPU |
| Facial-Attribute-Detection | QNN_DLC | float | Snapdragon® X2 Elite | 0.605 ms | 0 - 0 MB | NPU |
| Facial-Attribute-Detection | QNN_DLC | float | Snapdragon® X Elite | 0.98 ms | 0 - 0 MB | NPU |
| Facial-Attribute-Detection | QNN_DLC | float | Snapdragon® 8 Gen 3 Mobile | 0.639 ms | 0 - 82 MB | NPU |
| Facial-Attribute-Detection | QNN_DLC | float | Qualcomm® QCS8275 (Proxy) | 4.249 ms | 0 - 54 MB | NPU |
| Facial-Attribute-Detection | QNN_DLC | float | Qualcomm® QCS8550 (Proxy) | 0.851 ms | 0 - 101 MB | NPU |
| Facial-Attribute-Detection | QNN_DLC | float | Qualcomm® SA8775P | 1.395 ms | 0 - 56 MB | NPU |
| Facial-Attribute-Detection | QNN_DLC | float | Qualcomm® QCS9075 | 1.236 ms | 0 - 2 MB | NPU |
| Facial-Attribute-Detection | QNN_DLC | float | Qualcomm® QCS8450 (Proxy) | 1.361 ms | 0 - 78 MB | NPU |
| Facial-Attribute-Detection | QNN_DLC | float | Qualcomm® SA7255P | 4.249 ms | 0 - 54 MB | NPU |
| Facial-Attribute-Detection | QNN_DLC | float | Qualcomm® SA8295P | 1.568 ms | 0 - 48 MB | NPU |
| Facial-Attribute-Detection | QNN_DLC | float | Snapdragon® 8 Elite For Galaxy Mobile | 0.497 ms | 0 - 57 MB | NPU |
| Facial-Attribute-Detection | QNN_DLC | w8a8 | Snapdragon® 8 Elite Gen 5 Mobile | 0.211 ms | 0 - 49 MB | NPU |
| Facial-Attribute-Detection | QNN_DLC | w8a8 | Snapdragon® X2 Elite | 0.309 ms | 0 - 0 MB | NPU |
| Facial-Attribute-Detection | QNN_DLC | w8a8 | Snapdragon® X Elite | 0.509 ms | 0 - 0 MB | NPU |
| Facial-Attribute-Detection | QNN_DLC | w8a8 | Snapdragon® 8 Gen 3 Mobile | 0.281 ms | 0 - 67 MB | NPU |
| Facial-Attribute-Detection | QNN_DLC | w8a8 | Qualcomm® QCS6490 | 1.294 ms | 0 - 2 MB | NPU |
| Facial-Attribute-Detection | QNN_DLC | w8a8 | Qualcomm® QCS8275 (Proxy) | 1.074 ms | 0 - 47 MB | NPU |
| Facial-Attribute-Detection | QNN_DLC | w8a8 | Qualcomm® QCS8550 (Proxy) | 0.388 ms | 0 - 2 MB | NPU |
| Facial-Attribute-Detection | QNN_DLC | w8a8 | Qualcomm® SA8775P | 0.61 ms | 0 - 49 MB | NPU |
| Facial-Attribute-Detection | QNN_DLC | w8a8 | Qualcomm® QCS9075 | 0.492 ms | 0 - 2 MB | NPU |
| Facial-Attribute-Detection | QNN_DLC | w8a8 | Qualcomm® QCM6690 | 2.881 ms | 0 - 48 MB | NPU |
| Facial-Attribute-Detection | QNN_DLC | w8a8 | Qualcomm® QCS8450 (Proxy) | 0.627 ms | 0 - 64 MB | NPU |
| Facial-Attribute-Detection | QNN_DLC | w8a8 | Qualcomm® SA7255P | 1.074 ms | 0 - 47 MB | NPU |
| Facial-Attribute-Detection | QNN_DLC | w8a8 | Qualcomm® SA8295P | 0.842 ms | 0 - 45 MB | NPU |
| Facial-Attribute-Detection | QNN_DLC | w8a8 | Snapdragon® 8 Elite For Galaxy Mobile | 0.236 ms | 0 - 50 MB | NPU |
| Facial-Attribute-Detection | QNN_DLC | w8a8 | Snapdragon® 7 Gen 4 Mobile | 0.498 ms | 0 - 46 MB | NPU |
| Facial-Attribute-Detection | TFLITE | float | Snapdragon® 8 Elite Gen 5 Mobile | 0.412 ms | 0 - 76 MB | NPU |
| Facial-Attribute-Detection | TFLITE | float | Snapdragon® 8 Gen 3 Mobile | 0.606 ms | 0 - 98 MB | NPU |
| Facial-Attribute-Detection | TFLITE | float | Qualcomm® QCS8275 (Proxy) | 39.887 ms | 0 - 8 MB | CPU |
| Facial-Attribute-Detection | TFLITE | float | Qualcomm® QCS8550 (Proxy) | 0.817 ms | 0 - 30 MB | NPU |
| Facial-Attribute-Detection | TFLITE | float | Qualcomm® SA8775P | 1.381 ms | 0 - 73 MB | NPU |
| Facial-Attribute-Detection | TFLITE | float | Qualcomm® QCS9075 | 1.203 ms | 0 - 24 MB | NPU |
| Facial-Attribute-Detection | TFLITE | float | Qualcomm® QCS8450 (Proxy) | 1.304 ms | 0 - 95 MB | NPU |
| Facial-Attribute-Detection | TFLITE | float | Qualcomm® SA7255P | 39.887 ms | 0 - 8 MB | CPU |
| Facial-Attribute-Detection | TFLITE | float | Qualcomm® SA8295P | 1.564 ms | 0 - 63 MB | NPU |
| Facial-Attribute-Detection | TFLITE | float | Snapdragon® 8 Elite For Galaxy Mobile | 0.497 ms | 0 - 76 MB | NPU |
| Facial-Attribute-Detection | TFLITE | w8a8 | Snapdragon® 8 Elite Gen 5 Mobile | 0.202 ms | 0 - 50 MB | NPU |
| Facial-Attribute-Detection | TFLITE | w8a8 | Snapdragon® 8 Gen 3 Mobile | 0.288 ms | 0 - 65 MB | NPU |
| Facial-Attribute-Detection | TFLITE | w8a8 | Qualcomm® QCS6490 | 1.255 ms | 0 - 13 MB | NPU |
| Facial-Attribute-Detection | TFLITE | w8a8 | Qualcomm® QCS8275 (Proxy) | 1.088 ms | 0 - 47 MB | NPU |
| Facial-Attribute-Detection | TFLITE | w8a8 | Qualcomm® QCS8550 (Proxy) | 0.383 ms | 0 - 3 MB | NPU |
| Facial-Attribute-Detection | TFLITE | w8a8 | Qualcomm® SA8775P | 0.615 ms | 0 - 51 MB | NPU |
| Facial-Attribute-Detection | TFLITE | w8a8 | Qualcomm® QCS9075 | 0.483 ms | 0 - 13 MB | NPU |
| Facial-Attribute-Detection | TFLITE | w8a8 | Qualcomm® QCM6690 | 2.949 ms | 0 - 42 MB | NPU |
| Facial-Attribute-Detection | TFLITE | w8a8 | Qualcomm® QCS8450 (Proxy) | 0.611 ms | 0 - 64 MB | NPU |
| Facial-Attribute-Detection | TFLITE | w8a8 | Qualcomm® SA7255P | 1.088 ms | 0 - 47 MB | NPU |
| Facial-Attribute-Detection | TFLITE | w8a8 | Qualcomm® SA8295P | 0.826 ms | 0 - 43 MB | NPU |
| Facial-Attribute-Detection | TFLITE | w8a8 | Snapdragon® 8 Elite For Galaxy Mobile | 0.239 ms | 0 - 42 MB | NPU |
| Facial-Attribute-Detection | TFLITE | w8a8 | Snapdragon® 7 Gen 4 Mobile | 0.493 ms | 0 - 42 MB | NPU |
License
- The license for the original implementation of Facial-Attribute-Detection can be found here.
Community
- Join our AI Hub Slack community to collaborate, post questions and learn more about on-device AI.
- For questions or feedback please reach out to us.