---
library_name: pytorch
license: mit
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/v0.63.0/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 2.50, ONNX Runtime 1.27.1 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/pidnet/releases/v0.63.0/pidnet-onnx-float.zip)
| ONNX | w8a8 | Universal | QAIRT 2.50, ONNX Runtime 1.27.1 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/pidnet/releases/v0.63.0/pidnet-onnx-w8a8.zip)
| QNN_DLC | float | Universal | QAIRT 2.50 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/pidnet/releases/v0.63.0/pidnet-qnn_dlc-float.zip)
| QNN_DLC | w8a8 | Universal | QAIRT 2.50 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/pidnet/releases/v0.63.0/pidnet-qnn_dlc-w8a8.zip)
| TFLITE | float | Universal | QAIRT 2.50 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/pidnet/releases/v0.63.0/pidnet-tflite-float.zip)
| TFLITE | w8a8 | Universal | QAIRT 2.50 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/pidnet/releases/v0.63.0/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/v0.63.0/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/v0.63.0/src/qai_hub_models/models/pidnet) for usage instructions.

## Model Details

**Model Type:** Model_use_case.semantic_segmentation

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

## Performance Summary
| Model | Runtime | Precision | Chipset | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit
|---|---|---|---|---|---|---
| PidNet | ONNX | float | Snapdragon® 8 Elite Gen 5 For Galaxy Mobile | 16.389 ms | 5 - 242 MB | NPU
| PidNet | ONNX | float | Snapdragon® 8 Elite For Galaxy Mobile | 17.755 ms | 6 - 211 MB | NPU
| PidNet | ONNX | float | Snapdragon® X2 Elite | 13.292 ms | 24 - 24 MB | NPU
| PidNet | ONNX | float | Snapdragon® X Elite | 31.933 ms | 24 - 24 MB | NPU
| PidNet | ONNX | float | Snapdragon® 8 Gen 3 Mobile | 22.44 ms | 30 - 325 MB | NPU
| PidNet | ONNX | float | Snapdragon® 8 Gen 1 Mobile | 47.61 ms | 30 - 323 MB | NPU
| PidNet | ONNX | float | Qualcomm® Dragonwing™ IQ-8275 | 39.427 ms | 24 - 52 MB | NPU
| PidNet | ONNX | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 32.316 ms | 24 - 46 MB | NPU
| PidNet | ONNX | float | Qualcomm® QCS8450 | 47.61 ms | 30 - 323 MB | NPU
| PidNet | ONNX | float | Qualcomm® Dragonwing™ IQ-9075 | 41.808 ms | 24 - 51 MB | NPU
| PidNet | ONNX | float | Qualcomm® Dragonwing™ IQ-X7181 | 31.933 ms | 24 - 24 MB | NPU
| PidNet | ONNX | float | Qualcomm® Dragonwing™ Q-8750 | 17.755 ms | 6 - 211 MB | NPU
| PidNet | ONNX | w8a8 | Snapdragon® 8 Elite Gen 5 For Galaxy Mobile | 22.332 ms | 2 - 225 MB | NPU
| PidNet | ONNX | w8a8 | Snapdragon® 8 Elite For Galaxy Mobile | 41.075 ms | 1 - 209 MB | NPU
| PidNet | ONNX | w8a8 | Snapdragon® X2 Elite | 22.95 ms | 7 - 7 MB | NPU
| PidNet | ONNX | w8a8 | Snapdragon® X Elite | 54.452 ms | 7 - 7 MB | NPU
| PidNet | ONNX | w8a8 | Snapdragon® 8 Gen 3 Mobile | 38.501 ms | 7 - 261 MB | NPU
| PidNet | ONNX | w8a8 | Snapdragon® 8 Gen 1 Mobile | 54.859 ms | 8 - 263 MB | NPU
| PidNet | ONNX | w8a8 | Qualcomm® Dragonwing™ QCS6490 | 116.686 ms | 6 - 9 MB | NPU
| PidNet | ONNX | w8a8 | Qualcomm® Dragonwing™ IQ-8275 | 50.257 ms | 6 - 10 MB | NPU
| PidNet | ONNX | w8a8 | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 52.157 ms | 6 - 11 MB | NPU
| PidNet | ONNX | w8a8 | Qualcomm® QCS8450 | 54.859 ms | 8 - 263 MB | NPU
| PidNet | ONNX | w8a8 | Qualcomm® Dragonwing™ IQ-9075 | 54.412 ms | 6 - 9 MB | NPU
| PidNet | ONNX | w8a8 | Qualcomm® Dragonwing™ IQ-X7181 | 54.452 ms | 7 - 7 MB | NPU
| PidNet | ONNX | w8a8 | Qualcomm® Dragonwing™ Q-6690 | 236.789 ms | 5 - 219 MB | NPU
| PidNet | ONNX | w8a8 | Qualcomm® Dragonwing™ Q-7790 | 62.58 ms | 7 - 225 MB | NPU
| PidNet | ONNX | w8a8 | Qualcomm® Dragonwing™ Q-8750 | 41.075 ms | 1 - 209 MB | NPU
| PidNet | ONNX | w8a8 | Snapdragon® 7 Gen 4 Mobile | 62.58 ms | 7 - 225 MB | NPU
| PidNet | QNN_DLC | float | Snapdragon® 8 Elite Gen 5 For Galaxy Mobile | 12.535 ms | 23 - 275 MB | NPU
| PidNet | QNN_DLC | float | Snapdragon® 8 Elite For Galaxy Mobile | 18.331 ms | 11 - 242 MB | NPU
| PidNet | QNN_DLC | float | Snapdragon® X2 Elite | 14.229 ms | 24 - 24 MB | NPU
| PidNet | QNN_DLC | float | Snapdragon® X Elite | 38.012 ms | 24 - 24 MB | NPU
| PidNet | QNN_DLC | float | Snapdragon® 8 Gen 3 Mobile | 25.199 ms | 24 - 316 MB | NPU
| PidNet | QNN_DLC | float | Snapdragon® 8 Gen 1 Mobile | 50.953 ms | 4 - 297 MB | NPU
| PidNet | QNN_DLC | float | Qualcomm® Dragonwing™ IQ-8275 | 44.073 ms | 24 - 53 MB | NPU
| PidNet | QNN_DLC | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 36.425 ms | 24 - 26 MB | NPU
| PidNet | QNN_DLC | float | Qualcomm® SA8650P | 46.605 ms | 24 - 247 MB | NPU
| PidNet | QNN_DLC | float | Qualcomm® SA8255P | 46.605 ms | 24 - 247 MB | NPU
| PidNet | QNN_DLC | float | Qualcomm® QCS8450 | 50.953 ms | 4 - 297 MB | NPU
| PidNet | QNN_DLC | float | Qualcomm® Dragonwing™ IQ-9075 | 46.8 ms | 24 - 53 MB | NPU
| PidNet | QNN_DLC | float | Qualcomm® Dragonwing™ IQ-X7181 | 38.012 ms | 24 - 24 MB | NPU
| PidNet | QNN_DLC | float | Qualcomm® Dragonwing™ Q-8750 | 18.331 ms | 11 - 242 MB | NPU
| PidNet | QNN_DLC | float | Qualcomm® SA7255P | 116.892 ms | 24 - 231 MB | NPU
| PidNet | QNN_DLC | float | Qualcomm® SA8295P | 52.671 ms | 24 - 242 MB | NPU
| PidNet | QNN_DLC | w8a8 | Snapdragon® 8 Elite Gen 5 For Galaxy Mobile | 24.618 ms | 6 - 255 MB | NPU
| PidNet | QNN_DLC | w8a8 | Snapdragon® 8 Elite For Galaxy Mobile | 44.636 ms | 6 - 228 MB | NPU
| PidNet | QNN_DLC | w8a8 | Snapdragon® X2 Elite | 25.079 ms | 6 - 6 MB | NPU
| PidNet | QNN_DLC | w8a8 | Snapdragon® X Elite | 60.236 ms | 6 - 6 MB | NPU
| PidNet | QNN_DLC | w8a8 | Snapdragon® 8 Gen 3 Mobile | 43.073 ms | 6 - 256 MB | NPU
| PidNet | QNN_DLC | w8a8 | Snapdragon® 8 Gen 1 Mobile | 59.235 ms | 6 - 258 MB | NPU
| PidNet | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ IQ-8275 | 56.329 ms | 6 - 15 MB | NPU
| PidNet | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 57.392 ms | 6 - 17 MB | NPU
| PidNet | QNN_DLC | w8a8 | Qualcomm® QCS8450 | 59.235 ms | 6 - 258 MB | NPU
| PidNet | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ IQ-9075 | 59.532 ms | 6 - 15 MB | NPU
| PidNet | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ IQ-X7181 | 60.236 ms | 6 - 6 MB | NPU
| PidNet | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ Q-8750 | 44.636 ms | 6 - 228 MB | NPU
| PidNet | QNN_DLC | w8a8 | Qualcomm® SA7255P | 110.52 ms | 6 - 207 MB | NPU
| PidNet | QNN_DLC | w8a8 | Qualcomm® SA8295P | 66.502 ms | 6 - 212 MB | NPU
| PidNet | TFLITE | float | Snapdragon® 8 Elite Gen 5 For Galaxy Mobile | 12.693 ms | 2 - 255 MB | NPU
| PidNet | TFLITE | float | Snapdragon® 8 Elite For Galaxy Mobile | 18.622 ms | 2 - 238 MB | NPU
| PidNet | TFLITE | float | Snapdragon® 8 Gen 3 Mobile | 25.88 ms | 2 - 309 MB | NPU
| PidNet | TFLITE | float | Snapdragon® 8 Gen 1 Mobile | 49.327 ms | 0 - 312 MB | NPU
| PidNet | TFLITE | float | Qualcomm® Dragonwing™ IQ-8275 | 43.649 ms | 2 - 47 MB | NPU
| PidNet | TFLITE | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 36.102 ms | 2 - 5 MB | NPU
| PidNet | TFLITE | float | Qualcomm® SA8650P | 45.283 ms | 3 - 233 MB | NPU
| PidNet | TFLITE | float | Qualcomm® SA8255P | 45.283 ms | 3 - 233 MB | NPU
| PidNet | TFLITE | float | Qualcomm® QCS8450 | 49.327 ms | 0 - 312 MB | NPU
| PidNet | TFLITE | float | Qualcomm® Dragonwing™ IQ-9075 | 45.595 ms | 0 - 45 MB | NPU
| PidNet | TFLITE | float | Qualcomm® Dragonwing™ Q-8750 | 18.622 ms | 2 - 238 MB | NPU
| PidNet | TFLITE | float | Qualcomm® SA8295P | 52.299 ms | 3 - 233 MB | NPU
| PidNet | TFLITE | w8a8 | Snapdragon® 8 Elite Gen 5 For Galaxy Mobile | 22.097 ms | 1 - 250 MB | NPU
| PidNet | TFLITE | w8a8 | Snapdragon® 8 Elite For Galaxy Mobile | 69.04 ms | 1 - 222 MB | NPU
| PidNet | TFLITE | w8a8 | Snapdragon® 8 Gen 3 Mobile | 37.983 ms | 1 - 255 MB | NPU
| PidNet | TFLITE | w8a8 | Snapdragon® 8 Gen 1 Mobile | 50.971 ms | 1 - 255 MB | NPU
| PidNet | TFLITE | w8a8 | Qualcomm® Dragonwing™ QCS6490 | 182.211 ms | 2 - 73 MB | NPU
| PidNet | TFLITE | w8a8 | Qualcomm® Dragonwing™ IQ-8275 | 49.925 ms | 1 - 19 MB | NPU
| PidNet | TFLITE | w8a8 | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 50.202 ms | 1 - 3 MB | NPU
| PidNet | TFLITE | w8a8 | Qualcomm® SA8775P | 50.935 ms | 1 - 203 MB | NPU
| PidNet | TFLITE | w8a8 | Qualcomm® SA8650P | 50.935 ms | 1 - 203 MB | NPU
| PidNet | TFLITE | w8a8 | Qualcomm® SA8255P | 50.935 ms | 1 - 203 MB | NPU
| PidNet | TFLITE | w8a8 | Qualcomm® QCS8450 | 50.971 ms | 1 - 255 MB | NPU
| PidNet | TFLITE | w8a8 | Qualcomm® Dragonwing™ IQ-9075 | 51.615 ms | 0 - 17 MB | NPU
| PidNet | TFLITE | w8a8 | Qualcomm® Dragonwing™ Q-6690 | 242.224 ms | 2 - 187 MB | NPU
| PidNet | TFLITE | w8a8 | Qualcomm® Dragonwing™ Q-7790 | 66.091 ms | 2 - 209 MB | NPU
| PidNet | TFLITE | w8a8 | Qualcomm® Dragonwing™ Q-8750 | 69.04 ms | 1 - 222 MB | NPU
| PidNet | TFLITE | w8a8 | Qualcomm® SA7255P | 98.116 ms | 1 - 203 MB | NPU
| PidNet | TFLITE | w8a8 | Qualcomm® SA8295P | 57.79 ms | 1 - 207 MB | NPU
| PidNet | TFLITE | w8a8 | Snapdragon® 7 Gen 4 Mobile | 66.091 ms | 2 - 209 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).