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

---

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

# Person-Foot-Detection: Optimized for Qualcomm Devices

Real-time multiple person detection with accurate feet localization optimized for mobile and edge.

This is based on the implementation of Person-Foot-Detection found [here](https://github.com/qualcomm/ai-hub-models/blob/main/src/qai_hub_models/models/foot_track_net/model.py).
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/main/src/qai_hub_models/models/foot_track_net) 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 2.42, ONNX Runtime 1.25.0 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/foot_track_net/releases/v0.55.0/foot_track_net-onnx-float.zip)
| ONNX | w8a16 | Universal | QAIRT 2.42, ONNX Runtime 1.25.0 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/foot_track_net/releases/v0.55.0/foot_track_net-onnx-w8a16.zip)
| ONNX | w8a8 | Universal | QAIRT 2.42, ONNX Runtime 1.25.0 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/foot_track_net/releases/v0.55.0/foot_track_net-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/foot_track_net/releases/v0.55.0/foot_track_net-qnn_dlc-float.zip)
| QNN_DLC | w8a16 | Universal | QAIRT 2.45 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/foot_track_net/releases/v0.55.0/foot_track_net-qnn_dlc-w8a16.zip)
| QNN_DLC | w8a8 | Universal | QAIRT 2.45 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/foot_track_net/releases/v0.55.0/foot_track_net-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/foot_track_net/releases/v0.55.0/foot_track_net-tflite-float.zip)
| TFLITE | w8a8 | Universal | QAIRT 2.45 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/foot_track_net/releases/v0.55.0/foot_track_net-tflite-w8a8.zip)

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


### Option 2: Export with Custom Configurations

Use the [Qualcomm® AI Hub Models](https://github.com/qualcomm/ai-hub-models/blob/main/src/qai_hub_models/models/foot_track_net) 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 [Person-Foot-Detection on GitHub](https://github.com/qualcomm/ai-hub-models/blob/main/src/qai_hub_models/models/foot_track_net) for usage instructions.

## Model Details

**Model Type:** Model_use_case.object_detection

**Model Stats:**
- Inference latency: RealTime
- Input resolution: 640x480
- Number of output classes: 2
- Number of parameters: 2.53M
- Model size (float): 9.69 MB
- Model size (w8a8): 2.62 MB
- Model size (w8a16): 2.90 MB

## Performance Summary
| Model | Runtime | Precision | Chipset | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit
|---|---|---|---|---|---|---
| Person-Foot-Detection | ONNX | float | Snapdragon® 8 Elite Gen 5 Mobile | 2.34 ms | 0 - 160 MB | NPU
| Person-Foot-Detection | ONNX | float | Snapdragon® X2 Elite | 2.583 ms | 209 - 209 MB | NPU
| Person-Foot-Detection | ONNX | float | Snapdragon® X Elite | 4.587 ms | 177 - 177 MB | NPU
| Person-Foot-Detection | ONNX | float | Snapdragon® 8 Gen 3 Mobile | 3.277 ms | 3 - 189 MB | NPU
| Person-Foot-Detection | ONNX | float | Qualcomm® QCS8550 (Proxy) | 4.591 ms | 15 - 28 MB | NPU
| Person-Foot-Detection | ONNX | float | Snapdragon® 8 Elite For Galaxy Mobile | 2.822 ms | 0 - 156 MB | NPU
| Person-Foot-Detection | ONNX | float | Qualcomm® QCS9075 | 7.522 ms | 14 - 61 MB | NPU
| Person-Foot-Detection | ONNX | float | Qualcomm® QCS8750 | 2.822 ms | 0 - 156 MB | NPU
| Person-Foot-Detection | ONNX | float | Qualcomm® QCS7181 | 4.587 ms | 177 - 177 MB | NPU
| Person-Foot-Detection | ONNX | w8a16 | Snapdragon® 8 Elite Gen 5 Mobile | 2.448 ms | 0 - 174 MB | NPU
| Person-Foot-Detection | ONNX | w8a16 | Snapdragon® X2 Elite | 2.635 ms | 211 - 211 MB | NPU
| Person-Foot-Detection | ONNX | w8a16 | Snapdragon® X Elite | 4.014 ms | 178 - 178 MB | NPU
| Person-Foot-Detection | ONNX | w8a16 | Snapdragon® 8 Gen 3 Mobile | 2.745 ms | 0 - 199 MB | NPU
| Person-Foot-Detection | ONNX | w8a16 | Qualcomm® QCS6490 | 526.924 ms | 93 - 97 MB | CPU
| Person-Foot-Detection | ONNX | w8a16 | Qualcomm® QCS8550 (Proxy) | 3.93 ms | 0 - 46 MB | NPU
| Person-Foot-Detection | ONNX | w8a16 | Snapdragon® 7 Gen 4 Mobile | 272.017 ms | 95 - 103 MB | CPU
| Person-Foot-Detection | ONNX | w8a16 | Snapdragon® 8 Elite For Galaxy Mobile | 2.348 ms | 0 - 172 MB | NPU
| Person-Foot-Detection | ONNX | w8a16 | Qualcomm® QCM6690 | 286.449 ms | 95 - 103 MB | CPU
| Person-Foot-Detection | ONNX | w8a16 | Qualcomm® QCS9075 | 4.986 ms | 4 - 51 MB | NPU
| Person-Foot-Detection | ONNX | w8a16 | Qualcomm® QCS7790 | 272.017 ms | 95 - 103 MB | CPU
| Person-Foot-Detection | ONNX | w8a16 | Qualcomm® QCS8750 | 2.348 ms | 0 - 172 MB | NPU
| Person-Foot-Detection | ONNX | w8a16 | Qualcomm® QCS7181 | 4.014 ms | 178 - 178 MB | NPU
| Person-Foot-Detection | ONNX | w8a8 | Snapdragon® 8 Elite Gen 5 Mobile | 0.673 ms | 0 - 162 MB | NPU
| Person-Foot-Detection | ONNX | w8a8 | Snapdragon® X2 Elite | 0.766 ms | 211 - 211 MB | NPU
| Person-Foot-Detection | ONNX | w8a8 | Snapdragon® X Elite | 1.42 ms | 148 - 148 MB | NPU
| Person-Foot-Detection | ONNX | w8a8 | Snapdragon® 8 Gen 3 Mobile | 0.944 ms | 0 - 68 MB | NPU
| Person-Foot-Detection | ONNX | w8a8 | Qualcomm® QCS6490 | 76.322 ms | 50 - 58 MB | CPU
| Person-Foot-Detection | ONNX | w8a8 | Qualcomm® QCS8550 (Proxy) | 1.405 ms | 0 - 12 MB | NPU
| Person-Foot-Detection | ONNX | w8a8 | Qualcomm® QCS9075 | 1.79 ms | 0 - 47 MB | NPU
| Person-Foot-Detection | ONNX | w8a8 | Snapdragon® 7 Gen 4 Mobile | 58.984 ms | 51 - 61 MB | CPU
| Person-Foot-Detection | ONNX | w8a8 | Snapdragon® 8 Elite For Galaxy Mobile | 0.79 ms | 0 - 51 MB | NPU
| Person-Foot-Detection | ONNX | w8a8 | Qualcomm® QCM6690 | 75.488 ms | 51 - 61 MB | CPU
| Person-Foot-Detection | ONNX | w8a8 | Qualcomm® QCS7790 | 58.984 ms | 51 - 61 MB | CPU
| Person-Foot-Detection | ONNX | w8a8 | Qualcomm® QCS8750 | 0.79 ms | 0 - 51 MB | NPU
| Person-Foot-Detection | ONNX | w8a8 | Qualcomm® QCS7181 | 1.42 ms | 148 - 148 MB | NPU
| Person-Foot-Detection | QNN_DLC | float | Snapdragon® 8 Elite Gen 5 Mobile | 2.376 ms | 3 - 166 MB | NPU
| Person-Foot-Detection | QNN_DLC | float | Snapdragon® X2 Elite | 3.003 ms | 4 - 4 MB | NPU
| Person-Foot-Detection | QNN_DLC | float | Snapdragon® X Elite | 4.977 ms | 4 - 4 MB | NPU
| Person-Foot-Detection | QNN_DLC | float | Snapdragon® 8 Gen 3 Mobile | 3.208 ms | 4 - 185 MB | NPU
| Person-Foot-Detection | QNN_DLC | float | Qualcomm® QCS8550 (Proxy) | 4.458 ms | 4 - 5 MB | NPU
| Person-Foot-Detection | QNN_DLC | float | Qualcomm® SA8775P | 7.206 ms | 0 - 156 MB | NPU
| Person-Foot-Detection | QNN_DLC | float | Qualcomm® SA8650P | 7.206 ms | 0 - 156 MB | NPU
| Person-Foot-Detection | QNN_DLC | float | Qualcomm® SA8255P | 7.206 ms | 0 - 156 MB | NPU
| Person-Foot-Detection | QNN_DLC | float | Snapdragon® 8 Elite For Galaxy Mobile | 2.507 ms | 0 - 158 MB | NPU
| Person-Foot-Detection | QNN_DLC | float | Qualcomm® QCS8450 (Proxy) | 10.878 ms | 4 - 189 MB | NPU
| Person-Foot-Detection | QNN_DLC | float | Qualcomm® QCS9075 | 8.324 ms | 4 - 14 MB | NPU
| Person-Foot-Detection | QNN_DLC | float | Qualcomm® SA8295P | 7.984 ms | 0 - 156 MB | NPU
| Person-Foot-Detection | QNN_DLC | float | Qualcomm® QCS8750 | 2.507 ms | 0 - 158 MB | NPU
| Person-Foot-Detection | QNN_DLC | float | Qualcomm® QCS7181 | 4.977 ms | 4 - 4 MB | NPU
| Person-Foot-Detection | QNN_DLC | w8a16 | Snapdragon® 8 Elite Gen 5 Mobile | 1.912 ms | 2 - 167 MB | NPU
| Person-Foot-Detection | QNN_DLC | w8a16 | Snapdragon® X2 Elite | 2.613 ms | 2 - 2 MB | NPU
| Person-Foot-Detection | QNN_DLC | w8a16 | Snapdragon® 8 Gen 3 Mobile | 2.176 ms | 2 - 184 MB | NPU
| Person-Foot-Detection | QNN_DLC | w8a16 | Qualcomm® QCS6490 | 13.147 ms | 1 - 6 MB | NPU
| Person-Foot-Detection | QNN_DLC | w8a16 | Qualcomm® QCS8275 | 8.139 ms | 2 - 158 MB | NPU
| Person-Foot-Detection | QNN_DLC | w8a16 | Qualcomm® QCS8550 (Proxy) | 3.255 ms | 2 - 4 MB | NPU
| Person-Foot-Detection | QNN_DLC | w8a16 | Qualcomm® SA8775P | 3.833 ms | 2 - 162 MB | NPU
| Person-Foot-Detection | QNN_DLC | w8a16 | Qualcomm® SA8650P | 3.833 ms | 2 - 162 MB | NPU
| Person-Foot-Detection | QNN_DLC | w8a16 | Qualcomm® SA8255P | 3.833 ms | 2 - 162 MB | NPU
| Person-Foot-Detection | QNN_DLC | w8a16 | Snapdragon® 7 Gen 4 Mobile | 5.793 ms | 2 - 159 MB | NPU
| Person-Foot-Detection | QNN_DLC | w8a16 | Snapdragon® 8 Elite For Galaxy Mobile | 2.007 ms | 0 - 159 MB | NPU
| Person-Foot-Detection | QNN_DLC | w8a16 | Qualcomm® QCM6690 | 38.523 ms | 2 - 160 MB | NPU
| Person-Foot-Detection | QNN_DLC | w8a16 | Qualcomm® SA7255P | 8.139 ms | 2 - 158 MB | NPU
| Person-Foot-Detection | QNN_DLC | w8a16 | Qualcomm® SA8295P | 5.203 ms | 2 - 157 MB | NPU
| Person-Foot-Detection | QNN_DLC | w8a16 | Qualcomm® QCS9075 | 4.121 ms | 0 - 6 MB | NPU
| Person-Foot-Detection | QNN_DLC | w8a16 | Qualcomm® QCS8450 (Proxy) | 5.582 ms | 2 - 180 MB | NPU
| Person-Foot-Detection | QNN_DLC | w8a16 | Qualcomm® QCS7790 | 5.793 ms | 2 - 159 MB | NPU
| Person-Foot-Detection | QNN_DLC | w8a16 | Qualcomm® QCS8750 | 2.007 ms | 0 - 159 MB | NPU
| Person-Foot-Detection | QNN_DLC | w8a8 | Snapdragon® 8 Elite Gen 5 Mobile | 0.508 ms | 0 - 156 MB | NPU
| Person-Foot-Detection | QNN_DLC | w8a8 | Snapdragon® X2 Elite | 0.762 ms | 1 - 1 MB | NPU
| Person-Foot-Detection | QNN_DLC | w8a8 | Snapdragon® X Elite | 1.343 ms | 1 - 1 MB | NPU
| Person-Foot-Detection | QNN_DLC | w8a8 | Snapdragon® 8 Gen 3 Mobile | 0.769 ms | 1 - 53 MB | NPU
| Person-Foot-Detection | QNN_DLC | w8a8 | Qualcomm® QCS6490 | 5.348 ms | 0 - 4 MB | NPU
| Person-Foot-Detection | QNN_DLC | w8a8 | Qualcomm® QCS8275 | 3.55 ms | 1 - 39 MB | NPU
| Person-Foot-Detection | QNN_DLC | w8a8 | Qualcomm® QCS8550 (Proxy) | 1.153 ms | 1 - 2 MB | NPU
| Person-Foot-Detection | QNN_DLC | w8a8 | Qualcomm® SA8775P | 1.534 ms | 1 - 42 MB | NPU
| Person-Foot-Detection | QNN_DLC | w8a8 | Qualcomm® SA8650P | 1.534 ms | 1 - 42 MB | NPU
| Person-Foot-Detection | QNN_DLC | w8a8 | Qualcomm® SA8255P | 1.534 ms | 1 - 42 MB | NPU
| Person-Foot-Detection | QNN_DLC | w8a8 | Qualcomm® QCS9075 | 1.331 ms | 0 - 4 MB | NPU
| Person-Foot-Detection | QNN_DLC | w8a8 | Qualcomm® QCS8450 (Proxy) | 1.455 ms | 1 - 54 MB | NPU
| Person-Foot-Detection | QNN_DLC | w8a8 | Snapdragon® 7 Gen 4 Mobile | 1.781 ms | 1 - 152 MB | NPU
| Person-Foot-Detection | QNN_DLC | w8a8 | Snapdragon® 8 Elite For Galaxy Mobile | 0.642 ms | 0 - 155 MB | NPU
| Person-Foot-Detection | QNN_DLC | w8a8 | Qualcomm® QCM6690 | 14.653 ms | 1 - 151 MB | NPU
| Person-Foot-Detection | QNN_DLC | w8a8 | Qualcomm® SA8295P | 2.229 ms | 0 - 36 MB | NPU
| Person-Foot-Detection | QNN_DLC | w8a8 | Qualcomm® SA7255P | 3.55 ms | 1 - 39 MB | NPU
| Person-Foot-Detection | QNN_DLC | w8a8 | Qualcomm® QCS7790 | 1.781 ms | 1 - 152 MB | NPU
| Person-Foot-Detection | QNN_DLC | w8a8 | Qualcomm® QCS8750 | 0.642 ms | 0 - 155 MB | NPU
| Person-Foot-Detection | QNN_DLC | w8a8 | Qualcomm® QCS7181 | 1.343 ms | 1 - 1 MB | NPU
| Person-Foot-Detection | TFLITE | float | Snapdragon® 8 Elite Gen 5 Mobile | 2.249 ms | 0 - 159 MB | NPU
| Person-Foot-Detection | TFLITE | float | Snapdragon® 8 Gen 3 Mobile | 3.024 ms | 0 - 180 MB | NPU
| Person-Foot-Detection | TFLITE | float | Qualcomm® QCS8275 | 22.21 ms | 5 - 156 MB | NPU
| Person-Foot-Detection | TFLITE | float | Qualcomm® QCS8550 (Proxy) | 4.368 ms | 5 - 12 MB | NPU
| Person-Foot-Detection | TFLITE | float | Qualcomm® SA8775P | 6.993 ms | 5 - 161 MB | NPU
| Person-Foot-Detection | TFLITE | float | Qualcomm® SA8650P | 6.993 ms | 5 - 161 MB | NPU
| Person-Foot-Detection | TFLITE | float | Qualcomm® SA8255P | 6.993 ms | 5 - 161 MB | NPU
| Person-Foot-Detection | TFLITE | float | Snapdragon® 8 Elite For Galaxy Mobile | 2.682 ms | 0 - 152 MB | NPU
| Person-Foot-Detection | TFLITE | float | Qualcomm® QCS8450 (Proxy) | 8.689 ms | 5 - 188 MB | NPU
| Person-Foot-Detection | TFLITE | float | Qualcomm® QCS9075 | 7.537 ms | 4 - 19 MB | NPU
| Person-Foot-Detection | TFLITE | float | Qualcomm® SA8295P | 8.803 ms | 5 - 159 MB | NPU
| Person-Foot-Detection | TFLITE | float | Qualcomm® SA7255P | 22.21 ms | 5 - 156 MB | NPU
| Person-Foot-Detection | TFLITE | float | Qualcomm® QCS8750 | 2.682 ms | 0 - 152 MB | NPU
| Person-Foot-Detection | TFLITE | w8a8 | Snapdragon® 8 Elite Gen 5 Mobile | 0.512 ms | 0 - 154 MB | NPU
| Person-Foot-Detection | TFLITE | w8a8 | Snapdragon® 8 Gen 3 Mobile | 0.783 ms | 0 - 52 MB | NPU
| Person-Foot-Detection | TFLITE | w8a8 | Qualcomm® QCS6490 | 4.685 ms | 1 - 8 MB | NPU
| Person-Foot-Detection | TFLITE | w8a8 | Qualcomm® QCS8275 | 3.522 ms | 1 - 38 MB | NPU
| Person-Foot-Detection | TFLITE | w8a8 | Qualcomm® QCS8550 (Proxy) | 1.128 ms | 0 - 1 MB | NPU
| Person-Foot-Detection | TFLITE | w8a8 | Qualcomm® SA8775P | 1.561 ms | 0 - 40 MB | NPU
| Person-Foot-Detection | TFLITE | w8a8 | Qualcomm® SA8650P | 1.561 ms | 0 - 40 MB | NPU
| Person-Foot-Detection | TFLITE | w8a8 | Qualcomm® SA8255P | 1.561 ms | 0 - 40 MB | NPU
| Person-Foot-Detection | TFLITE | w8a8 | Qualcomm® QCS9075 | 1.366 ms | 0 - 7 MB | NPU
| Person-Foot-Detection | TFLITE | w8a8 | Qualcomm® QCS8450 (Proxy) | 1.471 ms | 0 - 52 MB | NPU
| Person-Foot-Detection | TFLITE | w8a8 | Snapdragon® 7 Gen 4 Mobile | 1.819 ms | 0 - 150 MB | NPU
| Person-Foot-Detection | TFLITE | w8a8 | Snapdragon® 8 Elite For Galaxy Mobile | 0.656 ms | 0 - 154 MB | NPU
| Person-Foot-Detection | TFLITE | w8a8 | Qualcomm® QCM6690 | 14.686 ms | 1 - 150 MB | NPU
| Person-Foot-Detection | TFLITE | w8a8 | Qualcomm® SA8295P | 2.237 ms | 0 - 35 MB | NPU
| Person-Foot-Detection | TFLITE | w8a8 | Qualcomm® SA7255P | 3.522 ms | 1 - 38 MB | NPU
| Person-Foot-Detection | TFLITE | w8a8 | Qualcomm® QCS7790 | 1.819 ms | 0 - 150 MB | NPU
| Person-Foot-Detection | TFLITE | w8a8 | Qualcomm® QCS8750 | 0.656 ms | 0 - 154 MB | NPU

## License
* The license for the original implementation of Person-Foot-Detection can be found
  [here](https://github.com/qualcomm/ai-hub-models/blob/main/LICENSE).

## References
* [Source Model Implementation](https://github.com/qualcomm/ai-hub-models/blob/main/src/qai_hub_models/models/foot_track_net/model.py)

## 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).