--- 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 Mobile Deployment ## Deep learning model for pixel-level semantic segmentation using pyramid pooling 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 provides scripts to run PSPNet on Qualcomm® devices. More details on model performance across various devices, can be found [here](https://aihub.qualcomm.com/models/pspnet). ### 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 | Model | Precision | Device | Chipset | Target Runtime | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit | Target Model |---|---|---|---|---|---|---|---|---| | PSPNet | float | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | TFLITE | 1668.832 ms | 117 - 878 MB | NPU | [PSPNet.tflite](https://huggingface.co/qualcomm/PSPNet/blob/main/PSPNet.tflite) | | PSPNet | float | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | QNN_DLC | 1501.062 ms | 0 - 681 MB | NPU | [PSPNet.dlc](https://huggingface.co/qualcomm/PSPNet/blob/main/PSPNet.dlc) | | PSPNet | float | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | TFLITE | 1517.133 ms | 15 - 533 MB | NPU | [PSPNet.tflite](https://huggingface.co/qualcomm/PSPNet/blob/main/PSPNet.tflite) | | PSPNet | float | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | QNN_DLC | 1115.893 ms | 3 - 400 MB | NPU | [PSPNet.dlc](https://huggingface.co/qualcomm/PSPNet/blob/main/PSPNet.dlc) | | PSPNet | float | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | TFLITE | 599.337 ms | 128 - 131 MB | NPU | [PSPNet.tflite](https://huggingface.co/qualcomm/PSPNet/blob/main/PSPNet.tflite) | | PSPNet | float | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | QNN_DLC | 593.125 ms | 3 - 5 MB | NPU | [PSPNet.dlc](https://huggingface.co/qualcomm/PSPNet/blob/main/PSPNet.dlc) | | PSPNet | float | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | ONNX | 1307.328 ms | 71 - 232 MB | NPU | [PSPNet.onnx.zip](https://huggingface.co/qualcomm/PSPNet/blob/main/PSPNet.onnx.zip) | | PSPNet | float | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | TFLITE | 3023.934 ms | 119 - 878 MB | NPU | [PSPNet.tflite](https://huggingface.co/qualcomm/PSPNet/blob/main/PSPNet.tflite) | | PSPNet | float | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | QNN_DLC | 646.417 ms | 0 - 682 MB | NPU | [PSPNet.dlc](https://huggingface.co/qualcomm/PSPNet/blob/main/PSPNet.dlc) | | PSPNet | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | TFLITE | 543.73 ms | 111 - 1069 MB | NPU | [PSPNet.tflite](https://huggingface.co/qualcomm/PSPNet/blob/main/PSPNet.tflite) | | PSPNet | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | QNN_DLC | 473.761 ms | 25 - 921 MB | NPU | [PSPNet.dlc](https://huggingface.co/qualcomm/PSPNet/blob/main/PSPNet.dlc) | | PSPNet | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | ONNX | 997.496 ms | 78 - 801 MB | NPU | [PSPNet.onnx.zip](https://huggingface.co/qualcomm/PSPNet/blob/main/PSPNet.onnx.zip) | | PSPNet | float | Samsung Galaxy S25 | Snapdragon® 8 Elite For Galaxy Mobile | TFLITE | 375.278 ms | 110 - 864 MB | NPU | [PSPNet.tflite](https://huggingface.co/qualcomm/PSPNet/blob/main/PSPNet.tflite) | | PSPNet | float | Samsung Galaxy S25 | Snapdragon® 8 Elite For Galaxy Mobile | QNN_DLC | 379.078 ms | 2 - 673 MB | NPU | [PSPNet.dlc](https://huggingface.co/qualcomm/PSPNet/blob/main/PSPNet.dlc) | | PSPNet | float | Samsung Galaxy S25 | Snapdragon® 8 Elite For Galaxy Mobile | ONNX | 975.577 ms | 138 - 716 MB | NPU | [PSPNet.onnx.zip](https://huggingface.co/qualcomm/PSPNet/blob/main/PSPNet.onnx.zip) | | PSPNet | float | Snapdragon 8 Elite Gen 5 QRD | Snapdragon® 8 Elite Gen5 Mobile | TFLITE | 320.19 ms | 67 - 829 MB | NPU | [PSPNet.tflite](https://huggingface.co/qualcomm/PSPNet/blob/main/PSPNet.tflite) | | PSPNet | float | Snapdragon 8 Elite Gen 5 QRD | Snapdragon® 8 Elite Gen5 Mobile | QNN_DLC | 382.858 ms | 5 - 695 MB | NPU | [PSPNet.dlc](https://huggingface.co/qualcomm/PSPNet/blob/main/PSPNet.dlc) | | PSPNet | float | Snapdragon 8 Elite Gen 5 QRD | Snapdragon® 8 Elite Gen5 Mobile | ONNX | 1048.343 ms | 13 - 604 MB | NPU | [PSPNet.onnx.zip](https://huggingface.co/qualcomm/PSPNet/blob/main/PSPNet.onnx.zip) | | PSPNet | float | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN_DLC | 601.791 ms | 3 - 3 MB | NPU | [PSPNet.dlc](https://huggingface.co/qualcomm/PSPNet/blob/main/PSPNet.dlc) | | PSPNet | float | Snapdragon X Elite CRD | Snapdragon® X Elite | ONNX | 1030.685 ms | 265 - 265 MB | NPU | [PSPNet.onnx.zip](https://huggingface.co/qualcomm/PSPNet/blob/main/PSPNet.onnx.zip) | ## Installation Install the package via pip: ```bash pip install qai-hub-models ``` ## Configure Qualcomm® AI Hub Workbench to run this model on a cloud-hosted device Sign-in to [Qualcomm® AI Hub Workbench](https://workbench.aihub.qualcomm.com/) with your Qualcomm® ID. Once signed in navigate to `Account -> Settings -> API Token`. With this API token, you can configure your client to run models on the cloud hosted devices. ```bash qai-hub configure --api_token API_TOKEN ``` Navigate to [docs](https://workbench.aihub.qualcomm.com/docs/) for more information. ## Demo off target The package contains a simple end-to-end demo that downloads pre-trained weights and runs this model on a sample input. ```bash python -m qai_hub_models.models.pspnet.demo ``` The above demo runs a reference implementation of pre-processing, model inference, and post processing. **NOTE**: If you want running in a Jupyter Notebook or Google Colab like environment, please add the following to your cell (instead of the above). ``` %run -m qai_hub_models.models.pspnet.demo ``` ### Run model on a cloud-hosted device In addition to the demo, you can also run the model on a cloud-hosted Qualcomm® device. This script does the following: * Performance check on-device on a cloud-hosted device * Downloads compiled assets that can be deployed on-device for Android. * Accuracy check between PyTorch and on-device outputs. ```bash python -m qai_hub_models.models.pspnet.export ``` ## How does this work? This [export script](https://aihub.qualcomm.com/models/pspnet/qai_hub_models/models/PSPNet/export.py) leverages [Qualcomm® AI Hub](https://aihub.qualcomm.com/) to optimize, validate, and deploy this model on-device. Lets go through each step below in detail: Step 1: **Compile model for on-device deployment** To compile a PyTorch model for on-device deployment, we first trace the model in memory using the `jit.trace` and then call the `submit_compile_job` API. ```python import torch import qai_hub as hub from qai_hub_models.models.pspnet import Model # Load the model torch_model = Model.from_pretrained() # Device device = hub.Device("Samsung Galaxy S25") # Trace model input_shape = torch_model.get_input_spec() sample_inputs = torch_model.sample_inputs() pt_model = torch.jit.trace(torch_model, [torch.tensor(data[0]) for _, data in sample_inputs.items()]) # Compile model on a specific device compile_job = hub.submit_compile_job( model=pt_model, device=device, input_specs=torch_model.get_input_spec(), ) # Get target model to run on-device target_model = compile_job.get_target_model() ``` Step 2: **Performance profiling on cloud-hosted device** After compiling models from step 1. Models can be profiled model on-device using the `target_model`. Note that this scripts runs the model on a device automatically provisioned in the cloud. Once the job is submitted, you can navigate to a provided job URL to view a variety of on-device performance metrics. ```python profile_job = hub.submit_profile_job( model=target_model, device=device, ) ``` Step 3: **Verify on-device accuracy** To verify the accuracy of the model on-device, you can run on-device inference on sample input data on the same cloud hosted device. ```python input_data = torch_model.sample_inputs() inference_job = hub.submit_inference_job( model=target_model, device=device, inputs=input_data, ) on_device_output = inference_job.download_output_data() ``` With the output of the model, you can compute like PSNR, relative errors or spot check the output with expected output. **Note**: This on-device profiling and inference requires access to Qualcomm® AI Hub Workbench. [Sign up for access](https://myaccount.qualcomm.com/signup). ## Run demo on a cloud-hosted device You can also run the demo on-device. ```bash python -m qai_hub_models.models.pspnet.demo --eval-mode on-device ``` **NOTE**: If you want running in a Jupyter Notebook or Google Colab like environment, please add the following to your cell (instead of the above). ``` %run -m qai_hub_models.models.pspnet.demo -- --eval-mode on-device ``` ## Deploying compiled model to Android The models can be deployed using multiple runtimes: - TensorFlow Lite (`.tflite` export): [This tutorial](https://www.tensorflow.org/lite/android/quickstart) provides a guide to deploy the .tflite model in an Android application. - QNN (`.so` export ): This [sample app](https://docs.qualcomm.com/bundle/publicresource/topics/80-63442-50/sample_app.html) provides instructions on how to use the `.so` shared library in an Android application. ## View on Qualcomm® AI Hub Get more details on PSPNet's performance across various devices [here](https://aihub.qualcomm.com/models/pspnet). Explore all available models on [Qualcomm® AI Hub](https://aihub.qualcomm.com/) ## 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).