PSPNet: Optimized for Qualcomm Devices
PSPNet (Pyramid Scene Parsing Network) is a semantic segmentation model that captures global context information by applying pyramid pooling modules. It is designed to improve scene understanding by aggregating contextual features at multiple scales.
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 1.24.1 | Download |
| ONNX | w8a8 | Universal | QAIRT 2.42, ONNX Runtime 1.24.1 | Download |
| QNN_DLC | float | Universal | QAIRT 2.43 | Download |
| QNN_DLC | w8a8 | Universal | QAIRT 2.43 | Download |
| TFLITE | float | Universal | QAIRT 2.43, TFLite 2.17.0 | Download |
| TFLITE | w8a8 | Universal | QAIRT 2.43, TFLite 2.17.0 | Download |
For more device-specific assets and performance metrics, visit PSPNet 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 PSPNet on GitHub for usage instructions.
Model Details
Model Type: Model_use_case.semantic_segmentation
Model Stats:
- Model checkpoint: pspnet101_ade20k.pth
- Input resolution: 1x3x473x473
- Number of parameters: 65.7M
- Model size (float): 251 MB
Performance Summary
| Model | Runtime | Precision | Chipset | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit |
|---|---|---|---|---|---|---|
| PSPNet | ONNX | float | Snapdragon® X Elite | 693.468 ms | 529 - 529 MB | NPU |
| PSPNet | ONNX | float | Snapdragon® 8 Gen 3 Mobile | 511.447 ms | 0 - 925 MB | NPU |
| PSPNet | ONNX | float | Qualcomm® QCS8550 (Proxy) | 607.007 ms | 1 - 163 MB | NPU |
| PSPNet | ONNX | float | Snapdragon® 8 Elite For Galaxy Mobile | 332.552 ms | 246 - 1003 MB | NPU |
| PSPNet | ONNX | float | Snapdragon® 8 Elite Gen 5 Mobile | 322.712 ms | 239 - 1006 MB | NPU |
| PSPNet | ONNX | float | Snapdragon® X2 Elite | 427.934 ms | 266 - 266 MB | NPU |
| PSPNet | ONNX | w8a8 | Snapdragon® X Elite | 28.578 ms | 133 - 133 MB | NPU |
| PSPNet | ONNX | w8a8 | Snapdragon® 8 Gen 3 Mobile | 22.041 ms | 69 - 489 MB | NPU |
| PSPNet | ONNX | w8a8 | Qualcomm® QCS6490 | 3300.144 ms | 196 - 274 MB | CPU |
| PSPNet | ONNX | w8a8 | Qualcomm® QCS8550 (Proxy) | 30.431 ms | 64 - 148 MB | NPU |
| PSPNet | ONNX | w8a8 | Qualcomm® QCS9075 | 34.296 ms | 70 - 74 MB | NPU |
| PSPNet | ONNX | w8a8 | Qualcomm® QCM6690 | 3080.851 ms | 83 - 93 MB | CPU |
| PSPNet | ONNX | w8a8 | Snapdragon® 8 Elite For Galaxy Mobile | 19.115 ms | 69 - 338 MB | NPU |
| PSPNet | ONNX | w8a8 | Snapdragon® 7 Gen 4 Mobile | 2608.347 ms | 50 - 62 MB | CPU |
| PSPNet | ONNX | w8a8 | Snapdragon® 8 Elite Gen 5 Mobile | 16.83 ms | 71 - 341 MB | NPU |
| PSPNet | ONNX | w8a8 | Snapdragon® X2 Elite | 15.152 ms | 131 - 131 MB | NPU |
| PSPNet | QNN_DLC | float | Snapdragon® X Elite | 546.735 ms | 3 - 3 MB | NPU |
| PSPNet | QNN_DLC | float | Snapdragon® 8 Gen 3 Mobile | 625.995 ms | 3 - 878 MB | NPU |
| PSPNet | QNN_DLC | float | Qualcomm® QCS8275 (Proxy) | 1865.874 ms | 3 - 722 MB | NPU |
| PSPNet | QNN_DLC | float | Qualcomm® QCS8550 (Proxy) | 604.157 ms | 3 - 5 MB | NPU |
| PSPNet | QNN_DLC | float | Qualcomm® QCS8450 (Proxy) | 1703.062 ms | 0 - 434 MB | NPU |
| PSPNet | QNN_DLC | float | Snapdragon® 8 Elite For Galaxy Mobile | 300.352 ms | 0 - 717 MB | NPU |
| PSPNet | QNN_DLC | float | Snapdragon® 8 Elite Gen 5 Mobile | 267.159 ms | 3 - 735 MB | NPU |
| PSPNet | QNN_DLC | float | Snapdragon® X2 Elite | 228.493 ms | 3 - 3 MB | NPU |
| PSPNet | QNN_DLC | w8a8 | Snapdragon® X Elite | 25.987 ms | 1 - 1 MB | NPU |
| PSPNet | QNN_DLC | w8a8 | Snapdragon® 8 Gen 3 Mobile | 20.557 ms | 0 - 364 MB | NPU |
| PSPNet | QNN_DLC | w8a8 | Qualcomm® QCS8275 (Proxy) | 63.815 ms | 1 - 262 MB | NPU |
| PSPNet | QNN_DLC | w8a8 | Qualcomm® QCS8550 (Proxy) | 27.447 ms | 0 - 19 MB | NPU |
| PSPNet | QNN_DLC | w8a8 | Qualcomm® QCS9075 | 31.726 ms | 1 - 35 MB | NPU |
| PSPNet | QNN_DLC | w8a8 | Qualcomm® QCS8450 (Proxy) | 37.2 ms | 1 - 357 MB | NPU |
| PSPNet | QNN_DLC | w8a8 | Snapdragon® 8 Elite For Galaxy Mobile | 18.188 ms | 1 - 233 MB | NPU |
| PSPNet | QNN_DLC | w8a8 | Snapdragon® 7 Gen 4 Mobile | 92.24 ms | 1 - 458 MB | NPU |
| PSPNet | QNN_DLC | w8a8 | Snapdragon® 8 Elite Gen 5 Mobile | 16.19 ms | 1 - 237 MB | NPU |
| PSPNet | QNN_DLC | w8a8 | Snapdragon® X2 Elite | 14.508 ms | 1 - 1 MB | NPU |
| PSPNet | TFLITE | float | Snapdragon® 8 Gen 3 Mobile | 529.045 ms | 0 - 1040 MB | NPU |
| PSPNet | TFLITE | float | Qualcomm® QCS8275 (Proxy) | 1684.21 ms | 114 - 948 MB | NPU |
| PSPNet | TFLITE | float | Qualcomm® QCS8550 (Proxy) | 612.52 ms | 128 - 131 MB | NPU |
| PSPNet | TFLITE | float | Qualcomm® QCS8450 (Proxy) | 1561.811 ms | 41 - 616 MB | NPU |
| PSPNet | TFLITE | float | Snapdragon® 8 Elite For Galaxy Mobile | 980.781 ms | 126 - 878 MB | NPU |
| PSPNet | TFLITE | float | Snapdragon® 8 Elite Gen 5 Mobile | 329.855 ms | 128 - 966 MB | NPU |
| PSPNet | TFLITE | w8a8 | Snapdragon® 8 Gen 3 Mobile | 21.554 ms | 28 - 416 MB | NPU |
| PSPNet | TFLITE | w8a8 | Qualcomm® QCS8275 (Proxy) | 66.698 ms | 32 - 303 MB | NPU |
| PSPNet | TFLITE | w8a8 | Qualcomm® QCS8550 (Proxy) | 29.027 ms | 32 - 34 MB | NPU |
| PSPNet | TFLITE | w8a8 | Qualcomm® QCS9075 | 32.058 ms | 32 - 131 MB | NPU |
| PSPNet | TFLITE | w8a8 | Qualcomm® QCS8450 (Proxy) | 40.822 ms | 32 - 415 MB | NPU |
| PSPNet | TFLITE | w8a8 | Snapdragon® 8 Elite For Galaxy Mobile | 19.445 ms | 30 - 283 MB | NPU |
| PSPNet | TFLITE | w8a8 | Snapdragon® 7 Gen 4 Mobile | 94.282 ms | 0 - 410 MB | NPU |
| PSPNet | TFLITE | w8a8 | Snapdragon® 8 Elite Gen 5 Mobile | 15.312 ms | 32 - 282 MB | NPU |
License
- The license for the original implementation of PSPNet 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.