PPE-Detection: Optimized for Qualcomm Devices
Detect if a person is wearing personal protective equipments (PPE) in real-time. This model's architecture was developed by Qualcomm. The model was trained by Qualcomm on a proprietary dataset, 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 2.37, ONNX Runtime 1.23.0 | Download |
| ONNX | w8a16 | Universal | QAIRT 2.37, ONNX Runtime 1.23.0 | Download |
| ONNX | w8a8 | Universal | QAIRT 2.37, ONNX Runtime 1.23.0 | Download |
| QNN_DLC | float | Universal | QAIRT 2.42 | Download |
| QNN_DLC | w8a16 | Universal | QAIRT 2.42 | Download |
| QNN_DLC | w8a8 | Universal | QAIRT 2.42 | Download |
| TFLITE | float | Universal | QAIRT 2.42, TFLite 2.17.0 | Download |
| TFLITE | w8a8 | Universal | QAIRT 2.42, TFLite 2.17.0 | Download |
For more device-specific assets and performance metrics, visit PPE-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 PPE-Detection on GitHub for usage instructions.
Model Details
Model Type: Model_use_case.object_detection
Model Stats:
- Inference latency: RealTime
- Input resolution: 320x192
- Number of output classes: 2
- Number of parameters: 6.19M
- Model size (float): 23.6 MB
- Model size (w8a8): 6.23 MB
- Model size (w8a16): 6.65 MB
Performance Summary
| Model | Runtime | Precision | Chipset | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit |
|---|---|---|---|---|---|---|
| PPE-Detection | ONNX | float | Snapdragon® 8 Gen 3 Mobile | 1.177 ms | 0 - 173 MB | NPU |
| PPE-Detection | ONNX | float | Qualcomm® QCS8550 (Proxy) | 1.579 ms | 0 - 23 MB | NPU |
| PPE-Detection | ONNX | float | Qualcomm® QCS9075 | 2.652 ms | 1 - 4 MB | NPU |
| PPE-Detection | ONNX | float | Snapdragon® 8 Elite For Galaxy Mobile | 0.865 ms | 0 - 127 MB | NPU |
| PPE-Detection | ONNX | float | Snapdragon® 8 Elite Gen 5 Mobile | 0.759 ms | 0 - 137 MB | NPU |
| PPE-Detection | ONNX | w8a16 | Snapdragon® X Elite | 1.12 ms | 6 - 6 MB | NPU |
| PPE-Detection | ONNX | w8a16 | Snapdragon® 8 Gen 3 Mobile | 0.826 ms | 0 - 134 MB | NPU |
| PPE-Detection | ONNX | w8a16 | Qualcomm® QCS6490 | 89.734 ms | 26 - 30 MB | CPU |
| PPE-Detection | ONNX | w8a16 | Qualcomm® QCS8550 (Proxy) | 1.157 ms | 0 - 128 MB | NPU |
| PPE-Detection | ONNX | w8a16 | Qualcomm® QCS9075 | 1.754 ms | 0 - 3 MB | NPU |
| PPE-Detection | ONNX | w8a16 | Qualcomm® QCM6690 | 51.845 ms | 26 - 36 MB | CPU |
| PPE-Detection | ONNX | w8a16 | Snapdragon® 8 Elite For Galaxy Mobile | 0.616 ms | 0 - 110 MB | NPU |
| PPE-Detection | ONNX | w8a16 | Snapdragon® 7 Gen 4 Mobile | 38.11 ms | 28 - 37 MB | CPU |
| PPE-Detection | ONNX | w8a16 | Snapdragon® 8 Elite Gen 5 Mobile | 0.54 ms | 0 - 109 MB | NPU |
| PPE-Detection | ONNX | w8a8 | Snapdragon® X Elite | 0.642 ms | 6 - 6 MB | NPU |
| PPE-Detection | ONNX | w8a8 | Snapdragon® 8 Gen 3 Mobile | 0.548 ms | 0 - 128 MB | NPU |
| PPE-Detection | ONNX | w8a8 | Qualcomm® QCS6490 | 16.289 ms | 6 - 17 MB | CPU |
| PPE-Detection | ONNX | w8a8 | Qualcomm® QCS8550 (Proxy) | 0.682 ms | 0 - 11 MB | NPU |
| PPE-Detection | ONNX | w8a8 | Qualcomm® QCS9075 | 1.056 ms | 0 - 3 MB | NPU |
| PPE-Detection | ONNX | w8a8 | Qualcomm® QCM6690 | 8.824 ms | 5 - 14 MB | CPU |
| PPE-Detection | ONNX | w8a8 | Snapdragon® 8 Elite For Galaxy Mobile | 0.422 ms | 0 - 108 MB | NPU |
| PPE-Detection | ONNX | w8a8 | Snapdragon® 7 Gen 4 Mobile | 6.686 ms | 7 - 16 MB | CPU |
| PPE-Detection | ONNX | w8a8 | Snapdragon® 8 Elite Gen 5 Mobile | 0.37 ms | 0 - 108 MB | NPU |
| PPE-Detection | QNN_DLC | float | Snapdragon® X Elite | 1.316 ms | 1 - 1 MB | NPU |
| PPE-Detection | QNN_DLC | float | Snapdragon® 8 Gen 3 Mobile | 0.797 ms | 0 - 90 MB | NPU |
| PPE-Detection | QNN_DLC | float | Qualcomm® QCS8275 (Proxy) | 4.872 ms | 1 - 69 MB | NPU |
| PPE-Detection | QNN_DLC | float | Qualcomm® QCS8550 (Proxy) | 1.128 ms | 1 - 2 MB | NPU |
| PPE-Detection | QNN_DLC | float | Qualcomm® SA8775P | 1.767 ms | 1 - 71 MB | NPU |
| PPE-Detection | QNN_DLC | float | Qualcomm® QCS9075 | 2.256 ms | 1 - 3 MB | NPU |
| PPE-Detection | QNN_DLC | float | Qualcomm® QCS8450 (Proxy) | 2.03 ms | 0 - 48 MB | NPU |
| PPE-Detection | QNN_DLC | float | Qualcomm® SA7255P | 4.872 ms | 1 - 69 MB | NPU |
| PPE-Detection | QNN_DLC | float | Qualcomm® SA8295P | 2.284 ms | 0 - 31 MB | NPU |
| PPE-Detection | QNN_DLC | float | Snapdragon® 8 Elite For Galaxy Mobile | 0.611 ms | 1 - 64 MB | NPU |
| PPE-Detection | QNN_DLC | float | Snapdragon® 8 Elite Gen 5 Mobile | 0.491 ms | 1 - 68 MB | NPU |
| PPE-Detection | QNN_DLC | w8a16 | Snapdragon® X Elite | 1.03 ms | 0 - 0 MB | NPU |
| PPE-Detection | QNN_DLC | w8a16 | Snapdragon® 8 Gen 3 Mobile | 0.57 ms | 0 - 56 MB | NPU |
| PPE-Detection | QNN_DLC | w8a16 | Qualcomm® QCS6490 | 2.987 ms | 0 - 2 MB | NPU |
| PPE-Detection | QNN_DLC | w8a16 | Qualcomm® QCS8275 (Proxy) | 2.114 ms | 0 - 34 MB | NPU |
| PPE-Detection | QNN_DLC | w8a16 | Qualcomm® QCS8550 (Proxy) | 0.857 ms | 0 - 2 MB | NPU |
| PPE-Detection | QNN_DLC | w8a16 | Qualcomm® SA8775P | 4.595 ms | 0 - 35 MB | NPU |
| PPE-Detection | QNN_DLC | w8a16 | Qualcomm® QCS9075 | 1.335 ms | 0 - 2 MB | NPU |
| PPE-Detection | QNN_DLC | w8a16 | Qualcomm® QCM6690 | 6.545 ms | 0 - 179 MB | NPU |
| PPE-Detection | QNN_DLC | w8a16 | Qualcomm® QCS8450 (Proxy) | 1.189 ms | 0 - 57 MB | NPU |
| PPE-Detection | QNN_DLC | w8a16 | Qualcomm® SA7255P | 2.114 ms | 0 - 34 MB | NPU |
| PPE-Detection | QNN_DLC | w8a16 | Qualcomm® SA8295P | 1.524 ms | 0 - 50 MB | NPU |
| PPE-Detection | QNN_DLC | w8a16 | Snapdragon® 8 Elite For Galaxy Mobile | 0.456 ms | 0 - 38 MB | NPU |
| PPE-Detection | QNN_DLC | w8a16 | Snapdragon® 7 Gen 4 Mobile | 0.979 ms | 0 - 51 MB | NPU |
| PPE-Detection | QNN_DLC | w8a16 | Snapdragon® 8 Elite Gen 5 Mobile | 0.381 ms | 0 - 39 MB | NPU |
| PPE-Detection | QNN_DLC | w8a8 | Snapdragon® X Elite | 0.602 ms | 0 - 0 MB | NPU |
| PPE-Detection | QNN_DLC | w8a8 | Snapdragon® 8 Gen 3 Mobile | 0.337 ms | 0 - 53 MB | NPU |
| PPE-Detection | QNN_DLC | w8a8 | Qualcomm® QCS6490 | 1.959 ms | 0 - 2 MB | NPU |
| PPE-Detection | QNN_DLC | w8a8 | Qualcomm® QCS8275 (Proxy) | 1.292 ms | 0 - 33 MB | NPU |
| PPE-Detection | QNN_DLC | w8a8 | Qualcomm® QCS8550 (Proxy) | 0.479 ms | 0 - 2 MB | NPU |
| PPE-Detection | QNN_DLC | w8a8 | Qualcomm® SA8775P | 0.693 ms | 0 - 36 MB | NPU |
| PPE-Detection | QNN_DLC | w8a8 | Qualcomm® QCS9075 | 0.758 ms | 0 - 2 MB | NPU |
| PPE-Detection | QNN_DLC | w8a8 | Qualcomm® QCM6690 | 3.316 ms | 0 - 178 MB | NPU |
| PPE-Detection | QNN_DLC | w8a8 | Qualcomm® QCS8450 (Proxy) | 0.781 ms | 0 - 53 MB | NPU |
| PPE-Detection | QNN_DLC | w8a8 | Qualcomm® SA7255P | 1.292 ms | 0 - 33 MB | NPU |
| PPE-Detection | QNN_DLC | w8a8 | Qualcomm® SA8295P | 0.984 ms | 0 - 30 MB | NPU |
| PPE-Detection | QNN_DLC | w8a8 | Snapdragon® 8 Elite For Galaxy Mobile | 0.273 ms | 0 - 33 MB | NPU |
| PPE-Detection | QNN_DLC | w8a8 | Snapdragon® 7 Gen 4 Mobile | 0.575 ms | 0 - 47 MB | NPU |
| PPE-Detection | QNN_DLC | w8a8 | Snapdragon® 8 Elite Gen 5 Mobile | 0.242 ms | 0 - 37 MB | NPU |
| PPE-Detection | TFLITE | float | Snapdragon® 8 Gen 3 Mobile | 0.804 ms | 0 - 91 MB | NPU |
| PPE-Detection | TFLITE | float | Qualcomm® QCS8275 (Proxy) | 4.91 ms | 0 - 72 MB | NPU |
| PPE-Detection | TFLITE | float | Qualcomm® QCS8550 (Proxy) | 1.125 ms | 0 - 10 MB | NPU |
| PPE-Detection | TFLITE | float | Qualcomm® SA8775P | 1.809 ms | 0 - 74 MB | NPU |
| PPE-Detection | TFLITE | float | Qualcomm® QCS9075 | 2.275 ms | 0 - 15 MB | NPU |
| PPE-Detection | TFLITE | float | Qualcomm® QCS8450 (Proxy) | 2.033 ms | 0 - 54 MB | NPU |
| PPE-Detection | TFLITE | float | Qualcomm® SA7255P | 4.91 ms | 0 - 72 MB | NPU |
| PPE-Detection | TFLITE | float | Qualcomm® SA8295P | 2.267 ms | 0 - 33 MB | NPU |
| PPE-Detection | TFLITE | float | Snapdragon® 8 Elite For Galaxy Mobile | 0.621 ms | 0 - 59 MB | NPU |
| PPE-Detection | TFLITE | float | Snapdragon® 8 Elite Gen 5 Mobile | 0.507 ms | 0 - 72 MB | NPU |
| PPE-Detection | TFLITE | w8a8 | Snapdragon® 8 Gen 3 Mobile | 0.4 ms | 0 - 51 MB | NPU |
| PPE-Detection | TFLITE | w8a8 | Qualcomm® QCS6490 | 4.11 ms | 0 - 17 MB | NPU |
| PPE-Detection | TFLITE | w8a8 | Qualcomm® QCS8275 (Proxy) | 1.436 ms | 0 - 31 MB | NPU |
| PPE-Detection | TFLITE | w8a8 | Qualcomm® QCS8550 (Proxy) | 0.565 ms | 0 - 2 MB | NPU |
| PPE-Detection | TFLITE | w8a8 | Qualcomm® SA8775P | 0.8 ms | 0 - 36 MB | NPU |
| PPE-Detection | TFLITE | w8a8 | Qualcomm® QCS9075 | 0.892 ms | 0 - 8 MB | NPU |
| PPE-Detection | TFLITE | w8a8 | Qualcomm® QCM6690 | 4.749 ms | 0 - 177 MB | NPU |
| PPE-Detection | TFLITE | w8a8 | Qualcomm® QCS8450 (Proxy) | 0.849 ms | 0 - 52 MB | NPU |
| PPE-Detection | TFLITE | w8a8 | Qualcomm® SA7255P | 1.436 ms | 0 - 31 MB | NPU |
| PPE-Detection | TFLITE | w8a8 | Qualcomm® SA8295P | 1.089 ms | 0 - 46 MB | NPU |
| PPE-Detection | TFLITE | w8a8 | Snapdragon® 8 Elite For Galaxy Mobile | 0.311 ms | 0 - 37 MB | NPU |
| PPE-Detection | TFLITE | w8a8 | Snapdragon® 7 Gen 4 Mobile | 1.324 ms | 0 - 46 MB | NPU |
| PPE-Detection | TFLITE | w8a8 | Snapdragon® 8 Elite Gen 5 Mobile | 0.272 ms | 0 - 38 MB | NPU |
License
- The license for the original implementation of PPE-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.
