---
library_name: pytorch
license: other
tags:
- real_time
- android
pipeline_tag: image-segmentation

---

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

# PidNet: Optimized for Qualcomm Devices

PIDNet (Proportional-Integral-Derivative Network) is a real-time semantic segmentation model based on  PID controllers

This is based on the implementation of PidNet found [here](https://github.com/XuJiacong/PIDNet).
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/pidnet) 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/pidnet/releases/v0.59.0/pidnet-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/pidnet/releases/v0.59.0/pidnet-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/pidnet/releases/v0.59.0/pidnet-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/pidnet/releases/v0.59.0/pidnet-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/pidnet/releases/v0.59.0/pidnet-tflite-float.zip)
| TFLITE | w8a8 | Universal | QAIRT 2.45 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/pidnet/releases/v0.59.0/pidnet-tflite-w8a8.zip)

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


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

## Model Details

**Model Type:** Model_use_case.semantic_segmentation

**Model Stats:**
- Model checkpoint: PIDNet_S_Cityscapes_val.pt
- Inference latency: RealTime
- Input resolution: 1024x2048
- Number of output classes: 19
- Number of parameters: 8.06M
- Model size (float): 29.1 MB
- Model size (w8a8): 8.02 MB

## Performance Summary
| Model | Runtime | Precision | Chipset | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit
|---|---|---|---|---|---|---
| PidNet | ONNX | float | Snapdragon® X2 Elite | 13.744 ms | 24 - 24 MB | NPU
| PidNet | ONNX | float | Snapdragon® X Elite | 32.283 ms | 24 - 24 MB | NPU
| PidNet | ONNX | float | Snapdragon® 8 Gen 3 Mobile | 23.087 ms | 30 - 331 MB | NPU
| PidNet | ONNX | float | Snapdragon® 8 Gen 1 Mobile | 55.313 ms | 26 - 332 MB | NPU
| PidNet | ONNX | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 32.647 ms | 24 - 46 MB | NPU
| PidNet | ONNX | float | Qualcomm® QCS8450 | 55.313 ms | 26 - 332 MB | NPU
| PidNet | ONNX | float | Qualcomm® Dragonwing™ IQ-9075 | 49.376 ms | 24 - 51 MB | NPU
| PidNet | ONNX | float | Qualcomm® Dragonwing™ IQ-X7181 | 32.283 ms | 24 - 24 MB | NPU
| PidNet | ONNX | float | Qualcomm® Dragonwing™ Q-8750 | 17.624 ms | 8 - 220 MB | NPU
| PidNet | ONNX | float | Snapdragon® 8 Elite Mobile | 17.624 ms | 8 - 220 MB | NPU
| PidNet | ONNX | float | Snapdragon® 8 Elite Gen 5 Mobile | 13.254 ms | 7 - 257 MB | NPU
| PidNet | ONNX | w8a8 | Snapdragon® X2 Elite | 25.929 ms | 7 - 7 MB | NPU
| PidNet | ONNX | w8a8 | Snapdragon® X Elite | 55.07 ms | 7 - 7 MB | NPU
| PidNet | ONNX | w8a8 | Snapdragon® 8 Gen 3 Mobile | 39.151 ms | 7 - 271 MB | NPU
| PidNet | ONNX | w8a8 | Snapdragon® 8 Gen 1 Mobile | 57.854 ms | 8 - 271 MB | NPU
| PidNet | ONNX | w8a8 | Qualcomm® Dragonwing™ QCS6490 | 158.684 ms | 6 - 9 MB | NPU
| PidNet | ONNX | w8a8 | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 52.548 ms | 6 - 12 MB | NPU
| PidNet | ONNX | w8a8 | Qualcomm® QCS8450 | 57.854 ms | 8 - 271 MB | NPU
| PidNet | ONNX | w8a8 | Qualcomm® Dragonwing™ IQ-9075 | 53.854 ms | 6 - 9 MB | NPU
| PidNet | ONNX | w8a8 | Qualcomm® Dragonwing™ IQ-X7181 | 55.07 ms | 7 - 7 MB | NPU
| PidNet | ONNX | w8a8 | Qualcomm® Dragonwing™ Q-8750 | 45.21 ms | 2 - 218 MB | NPU
| PidNet | ONNX | w8a8 | Snapdragon® 8 Elite Mobile | 45.21 ms | 2 - 218 MB | NPU
| PidNet | ONNX | w8a8 | Snapdragon® 8 Elite Gen 5 Mobile | 25.835 ms | 2 - 237 MB | NPU
| PidNet | QNN_DLC | float | Snapdragon® X2 Elite | 13.648 ms | 24 - 24 MB | NPU
| PidNet | QNN_DLC | float | Snapdragon® X Elite | 38.863 ms | 24 - 24 MB | NPU
| PidNet | QNN_DLC | float | Snapdragon® 8 Gen 3 Mobile | 25.44 ms | 15 - 321 MB | NPU
| PidNet | QNN_DLC | float | Snapdragon® 8 Gen 1 Mobile | 75.769 ms | 4 - 310 MB | NPU
| PidNet | QNN_DLC | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 37.314 ms | 24 - 306 MB | NPU
| PidNet | QNN_DLC | float | Qualcomm® SA8775P | 46.572 ms | 24 - 243 MB | NPU
| PidNet | QNN_DLC | float | Qualcomm® SA8650P | 46.572 ms | 24 - 243 MB | NPU
| PidNet | QNN_DLC | float | Qualcomm® SA8255P | 46.572 ms | 24 - 243 MB | NPU
| PidNet | QNN_DLC | float | Qualcomm® QCS8450 | 75.769 ms | 4 - 310 MB | NPU
| PidNet | QNN_DLC | float | Qualcomm® Dragonwing™ IQ-9075 | 47.54 ms | 26 - 54 MB | NPU
| PidNet | QNN_DLC | float | Qualcomm® Dragonwing™ IQ-X7181 | 38.863 ms | 24 - 24 MB | NPU
| PidNet | QNN_DLC | float | Qualcomm® Dragonwing™ Q-8750 | 18.3 ms | 23 - 268 MB | NPU
| PidNet | QNN_DLC | float | Qualcomm® SA8295P | 51.999 ms | 24 - 254 MB | NPU
| PidNet | QNN_DLC | float | Snapdragon® 8 Elite Mobile | 18.3 ms | 23 - 268 MB | NPU
| PidNet | QNN_DLC | float | Snapdragon® 8 Elite Gen 5 Mobile | 12.133 ms | 16 - 288 MB | NPU
| PidNet | QNN_DLC | w8a8 | Snapdragon® X2 Elite | 26.231 ms | 6 - 6 MB | NPU
| PidNet | QNN_DLC | w8a8 | Snapdragon® X Elite | 60.559 ms | 6 - 6 MB | NPU
| PidNet | QNN_DLC | w8a8 | Snapdragon® 8 Gen 3 Mobile | 43.014 ms | 6 - 268 MB | NPU
| PidNet | QNN_DLC | w8a8 | Snapdragon® 8 Gen 1 Mobile | 63.841 ms | 6 - 268 MB | NPU
| PidNet | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ QCS8275 | 111.372 ms | 6 - 218 MB | NPU
| PidNet | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 57.836 ms | 6 - 9 MB | NPU
| PidNet | QNN_DLC | w8a8 | Qualcomm® SA8775P | 58.506 ms | 6 - 219 MB | NPU
| PidNet | QNN_DLC | w8a8 | Qualcomm® SA8650P | 58.506 ms | 6 - 219 MB | NPU
| PidNet | QNN_DLC | w8a8 | Qualcomm® SA8255P | 58.506 ms | 6 - 219 MB | NPU
| PidNet | QNN_DLC | w8a8 | Qualcomm® QCS8450 | 63.841 ms | 6 - 268 MB | NPU
| PidNet | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ IQ-9075 | 66.963 ms | 4 - 13 MB | NPU
| PidNet | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ IQ-X7181 | 60.559 ms | 6 - 6 MB | NPU
| PidNet | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ Q-8750 | 43.862 ms | 6 - 243 MB | NPU
| PidNet | QNN_DLC | w8a8 | Qualcomm® SA7255P | 111.372 ms | 6 - 218 MB | NPU
| PidNet | QNN_DLC | w8a8 | Qualcomm® SA8295P | 66.436 ms | 6 - 223 MB | NPU
| PidNet | QNN_DLC | w8a8 | Snapdragon® 8 Elite Mobile | 43.862 ms | 6 - 243 MB | NPU
| PidNet | QNN_DLC | w8a8 | Snapdragon® 8 Elite Gen 5 Mobile | 24.603 ms | 6 - 272 MB | NPU
| PidNet | TFLITE | float | Snapdragon® 8 Gen 3 Mobile | 24.998 ms | 1 - 320 MB | NPU
| PidNet | TFLITE | float | Snapdragon® 8 Gen 1 Mobile | 74.636 ms | 0 - 325 MB | NPU
| PidNet | TFLITE | float | Qualcomm® Dragonwing™ QCS8275 | 116.584 ms | 3 - 231 MB | NPU
| PidNet | TFLITE | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 36.628 ms | 2 - 25 MB | NPU
| PidNet | TFLITE | float | Qualcomm® SA8775P | 45.335 ms | 0 - 229 MB | NPU
| PidNet | TFLITE | float | Qualcomm® SA8650P | 45.335 ms | 0 - 229 MB | NPU
| PidNet | TFLITE | float | Qualcomm® SA8255P | 45.335 ms | 0 - 229 MB | NPU
| PidNet | TFLITE | float | Qualcomm® QCS8450 | 74.636 ms | 0 - 325 MB | NPU
| PidNet | TFLITE | float | Qualcomm® Dragonwing™ IQ-9075 | 46.698 ms | 0 - 45 MB | NPU
| PidNet | TFLITE | float | Qualcomm® Dragonwing™ Q-8750 | 18.786 ms | 2 - 252 MB | NPU
| PidNet | TFLITE | float | Qualcomm® SA7255P | 116.584 ms | 3 - 231 MB | NPU
| PidNet | TFLITE | float | Qualcomm® SA8295P | 51.012 ms | 3 - 245 MB | NPU
| PidNet | TFLITE | float | Snapdragon® 8 Elite Mobile | 18.786 ms | 2 - 252 MB | NPU
| PidNet | TFLITE | float | Snapdragon® 8 Elite Gen 5 Mobile | 12.867 ms | 2 - 273 MB | NPU
| PidNet | TFLITE | w8a8 | Snapdragon® 8 Gen 3 Mobile | 38.042 ms | 0 - 265 MB | NPU
| PidNet | TFLITE | w8a8 | Snapdragon® 8 Gen 1 Mobile | 58.248 ms | 1 - 268 MB | NPU
| PidNet | TFLITE | w8a8 | Qualcomm® Dragonwing™ QCS6490 | 199.024 ms | 3 - 72 MB | NPU
| PidNet | TFLITE | w8a8 | Qualcomm® Dragonwing™ QCS8275 | 98.245 ms | 1 - 213 MB | NPU
| PidNet | TFLITE | w8a8 | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 50.266 ms | 1 - 3 MB | NPU
| PidNet | TFLITE | w8a8 | Qualcomm® SA8775P | 51.042 ms | 1 - 214 MB | NPU
| PidNet | TFLITE | w8a8 | Qualcomm® SA8650P | 51.042 ms | 1 - 214 MB | NPU
| PidNet | TFLITE | w8a8 | Qualcomm® SA8255P | 51.042 ms | 1 - 214 MB | NPU
| PidNet | TFLITE | w8a8 | Qualcomm® QCS8450 | 58.248 ms | 1 - 268 MB | NPU
| PidNet | TFLITE | w8a8 | Qualcomm® Dragonwing™ IQ-9075 | 51.634 ms | 1 - 17 MB | NPU
| PidNet | TFLITE | w8a8 | Qualcomm® Dragonwing™ Q-6690 | 242.962 ms | 3 - 235 MB | NPU
| PidNet | TFLITE | w8a8 | Qualcomm® Dragonwing™ Q-7790 | 65.959 ms | 3 - 221 MB | NPU
| PidNet | TFLITE | w8a8 | Qualcomm® Dragonwing™ Q-8750 | 69.809 ms | 1 - 234 MB | NPU
| PidNet | TFLITE | w8a8 | Qualcomm® SA7255P | 98.245 ms | 1 - 213 MB | NPU
| PidNet | TFLITE | w8a8 | Qualcomm® SA8295P | 58.059 ms | 1 - 219 MB | NPU
| PidNet | TFLITE | w8a8 | Snapdragon® 8 Elite Mobile | 69.809 ms | 1 - 234 MB | NPU
| PidNet | TFLITE | w8a8 | Snapdragon® 8 Elite Gen 5 Mobile | 22.237 ms | 0 - 267 MB | NPU
| PidNet | TFLITE | w8a8 | Snapdragon® 7 Gen 4 Mobile | 65.959 ms | 3 - 221 MB | NPU

## License
* The license for the original implementation of PidNet can be found
  [here](https://github.com/XuJiacong/PIDNet/blob/main/LICENSE).

## References
* [PIDNet A Real-time Semantic Segmentation Network Inspired from PID Controller Segmentation of Road Scenes](https://arxiv.org/abs/2206.02066)
* [Source Model Implementation](https://github.com/XuJiacong/PIDNet)

## 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).