--- library_name: pytorch license: other tags: - bu_auto - android pipeline_tag: image-segmentation --- ![](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/pspnet/web-assets/model_demo.png) # 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](https://github.com/qualcomm/ai-hub-models/blob/main/src/qai_hub_models/models/pspnet) 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 1.24.3 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/pspnet/releases/v0.51.0/pspnet-onnx-float.zip) | ONNX | w8a8 | Universal | QAIRT 2.42, ONNX Runtime 1.24.3 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/pspnet/releases/v0.51.0/pspnet-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/pspnet/releases/v0.51.0/pspnet-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/pspnet/releases/v0.51.0/pspnet-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/pspnet/releases/v0.51.0/pspnet-tflite-float.zip) | TFLITE | w8a8 | Universal | QAIRT 2.45 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/pspnet/releases/v0.51.0/pspnet-tflite-w8a8.zip) For more device-specific assets and performance metrics, visit **[PSPNet on Qualcomm® AI Hub](https://aihub.qualcomm.com/models/pspnet)**. ### 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/pspnet) 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](https://github.com/qualcomm/ai-hub-models/blob/main/src/qai_hub_models/models/pspnet) 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® 8 Elite Gen 5 Mobile | 323.183 ms | 239 - 1007 MB | NPU | PSPNet | ONNX | float | Snapdragon® X2 Elite | 429.565 ms | 266 - 266 MB | NPU | PSPNet | ONNX | float | Snapdragon® X Elite | 695.088 ms | 529 - 529 MB | NPU | PSPNet | ONNX | float | Snapdragon® 8 Gen 3 Mobile | 505.529 ms | 148 - 1065 MB | NPU | PSPNet | ONNX | float | Qualcomm® QCS8550 (Proxy) | 604.578 ms | 0 - 680 MB | NPU | PSPNet | ONNX | float | Qualcomm® QCS9075 | 1758.419 ms | 78 - 84 MB | NPU | PSPNet | ONNX | float | Snapdragon® 8 Elite For Galaxy Mobile | 333.069 ms | 245 - 1002 MB | NPU | PSPNet | ONNX | w8a8 | Snapdragon® 8 Elite Gen 5 Mobile | 17.157 ms | 72 - 342 MB | NPU | PSPNet | ONNX | w8a8 | Snapdragon® X2 Elite | 15.122 ms | 131 - 131 MB | NPU | PSPNet | ONNX | w8a8 | Snapdragon® X Elite | 28.639 ms | 133 - 133 MB | NPU | PSPNet | ONNX | w8a8 | Snapdragon® 8 Gen 3 Mobile | 21.995 ms | 72 - 492 MB | NPU | PSPNet | ONNX | w8a8 | Qualcomm® QCS6490 | 3430.903 ms | 204 - 283 MB | CPU | PSPNet | ONNX | w8a8 | Qualcomm® QCS8550 (Proxy) | 30.172 ms | 64 - 147 MB | NPU | PSPNet | ONNX | w8a8 | Qualcomm® QCS9075 | 34.249 ms | 70 - 74 MB | NPU | PSPNet | ONNX | w8a8 | Qualcomm® QCM6690 | 3282.281 ms | 53 - 64 MB | CPU | PSPNet | ONNX | w8a8 | Snapdragon® 8 Elite For Galaxy Mobile | 19.141 ms | 72 - 343 MB | NPU | PSPNet | ONNX | w8a8 | Snapdragon® 7 Gen 4 Mobile | 2610.165 ms | 64 - 76 MB | CPU | PSPNet | QNN_DLC | float | Snapdragon® 8 Elite Gen 5 Mobile | 269.515 ms | 7 - 752 MB | NPU | PSPNet | QNN_DLC | float | Snapdragon® X2 Elite | 228.12 ms | 3 - 3 MB | NPU | PSPNet | QNN_DLC | float | Snapdragon® X Elite | 530.704 ms | 3 - 3 MB | NPU | PSPNet | QNN_DLC | float | Snapdragon® 8 Gen 3 Mobile | 405.562 ms | 0 - 945 MB | NPU | PSPNet | QNN_DLC | float | Qualcomm® QCS8275 (Proxy) | 1330.613 ms | 2 - 725 MB | NPU | PSPNet | QNN_DLC | float | Qualcomm® QCS8550 (Proxy) | 589.618 ms | 3 - 13 MB | NPU | PSPNet | QNN_DLC | float | Qualcomm® QCS9075 | 1760.21 ms | 3 - 135 MB | NPU | PSPNet | QNN_DLC | float | Qualcomm® QCS8450 (Proxy) | 1700.398 ms | 0 - 441 MB | NPU | PSPNet | QNN_DLC | float | Snapdragon® 8 Elite For Galaxy Mobile | 302.547 ms | 0 - 725 MB | NPU | PSPNet | QNN_DLC | w8a8 | Snapdragon® 8 Elite Gen 5 Mobile | 15.395 ms | 1 - 310 MB | NPU | PSPNet | QNN_DLC | w8a8 | Snapdragon® X2 Elite | 14.501 ms | 1 - 1 MB | NPU | PSPNet | QNN_DLC | w8a8 | Snapdragon® X Elite | 26.163 ms | 1 - 1 MB | NPU | PSPNet | QNN_DLC | w8a8 | Snapdragon® 8 Gen 3 Mobile | 20.357 ms | 1 - 357 MB | NPU | PSPNet | QNN_DLC | w8a8 | Qualcomm® QCS8275 (Proxy) | 63.842 ms | 1 - 259 MB | NPU | PSPNet | QNN_DLC | w8a8 | Qualcomm® QCS8550 (Proxy) | 28.87 ms | 1 - 11 MB | NPU | PSPNet | QNN_DLC | w8a8 | Qualcomm® QCS9075 | 31.865 ms | 1 - 35 MB | NPU | PSPNet | QNN_DLC | w8a8 | Qualcomm® QCS8450 (Proxy) | 37.388 ms | 1 - 352 MB | NPU | PSPNet | QNN_DLC | w8a8 | Snapdragon® 8 Elite For Galaxy Mobile | 18.074 ms | 1 - 229 MB | NPU | PSPNet | QNN_DLC | w8a8 | Snapdragon® 7 Gen 4 Mobile | 90.511 ms | 1 - 417 MB | NPU | PSPNet | TFLITE | float | Snapdragon® 8 Elite Gen 5 Mobile | 324.12 ms | 128 - 961 MB | NPU | PSPNet | TFLITE | float | Snapdragon® 8 Gen 3 Mobile | 514.959 ms | 126 - 1257 MB | NPU | PSPNet | TFLITE | float | Qualcomm® QCS8275 (Proxy) | 1686.092 ms | 109 - 949 MB | NPU | PSPNet | TFLITE | float | Qualcomm® QCS8550 (Proxy) | 598.496 ms | 120 - 122 MB | NPU | PSPNet | TFLITE | float | Qualcomm® QCS9075 | 1766.366 ms | 0 - 271 MB | NPU | PSPNet | TFLITE | float | Qualcomm® QCS8450 (Proxy) | 1583.889 ms | 90 - 688 MB | NPU | PSPNet | TFLITE | float | Snapdragon® 8 Elite For Galaxy Mobile | 988.643 ms | 64 - 878 MB | NPU | PSPNet | TFLITE | w8a8 | Snapdragon® 8 Elite Gen 5 Mobile | 17.313 ms | 32 - 323 MB | NPU | PSPNet | TFLITE | w8a8 | Snapdragon® 8 Gen 3 Mobile | 26.198 ms | 32 - 416 MB | NPU | PSPNet | TFLITE | w8a8 | Qualcomm® QCS8275 (Proxy) | 74.504 ms | 32 - 301 MB | NPU | PSPNet | TFLITE | w8a8 | Qualcomm® QCS8550 (Proxy) | 35.741 ms | 32 - 34 MB | NPU | PSPNet | TFLITE | w8a8 | Qualcomm® QCS9075 | 37.769 ms | 32 - 131 MB | NPU | PSPNet | TFLITE | w8a8 | Qualcomm® QCS8450 (Proxy) | 54.778 ms | 32 - 414 MB | NPU | PSPNet | TFLITE | w8a8 | Snapdragon® 8 Elite For Galaxy Mobile | 22.088 ms | 32 - 283 MB | NPU | PSPNet | TFLITE | w8a8 | Snapdragon® 7 Gen 4 Mobile | 357.743 ms | 384 - 905 MB | NPU ## License * The license for the original implementation of PSPNet can be found [here](https://github.com/hszhao/semseg/blob/master/LICENSE). ## 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).