--- library_name: pytorch license: apache-2.0 tags: - android pipeline_tag: keypoint-detection --- ![](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/litehrnet/web-assets/model_demo.png) # LiteHRNet: Optimized for Qualcomm Devices LiteHRNet is a machine learning model that detects human pose and returns a location and confidence for each of 17 joints. This is based on the implementation of LiteHRNet found [here](https://github.com/HRNet/Lite-HRNet). 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/litehrnet) 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/litehrnet/releases/v0.63.0/litehrnet-onnx-float.zip) | QNN_DLC | float | Universal | QAIRT 2.50 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/litehrnet/releases/v0.63.0/litehrnet-qnn_dlc-float.zip) | TFLITE | float | Universal | QAIRT 2.50 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/litehrnet/releases/v0.63.0/litehrnet-tflite-float.zip) For more device-specific assets and performance metrics, visit **[LiteHRNet on Qualcomm® AI Hub](https://aihub.qualcomm.com/models/litehrnet)**. ### 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/litehrnet) 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 [LiteHRNet on GitHub](https://github.com/qualcomm/ai-hub-models/blob/v0.63.0/src/qai_hub_models/models/litehrnet) for usage instructions. ## Model Details **Model Type:** Model_use_case.pose_estimation **Model Stats:** - Input resolution: 256x192 - Model size (float): 4.49 MB - Number of parameters: 1.11M ## Performance Summary | Model | Runtime | Precision | Chipset | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit |---|---|---|---|---|---|--- | LiteHRNet | ONNX | float | Snapdragon® 8 Elite Gen 5 For Galaxy Mobile | 2.505 ms | 0 - 102 MB | NPU | LiteHRNet | ONNX | float | Snapdragon® 8 Elite For Galaxy Mobile | 2.669 ms | 0 - 100 MB | NPU | LiteHRNet | ONNX | float | Snapdragon® X2 Elite | 2.61 ms | 2 - 2 MB | NPU | LiteHRNet | ONNX | float | Snapdragon® X Elite | 5.636 ms | 5 - 5 MB | NPU | LiteHRNet | ONNX | float | Snapdragon® 8 Gen 3 Mobile | 3.029 ms | 0 - 126 MB | NPU | LiteHRNet | ONNX | float | Snapdragon® 8 Gen 1 Mobile | 6.298 ms | 1 - 122 MB | NPU | LiteHRNet | ONNX | float | Qualcomm® Dragonwing™ IQ-8275 | 4.375 ms | 1 - 5 MB | NPU | LiteHRNet | ONNX | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 5.335 ms | 0 - 8 MB | NPU | LiteHRNet | ONNX | float | Qualcomm® QCS8450 | 6.298 ms | 1 - 122 MB | NPU | LiteHRNet | ONNX | float | Qualcomm® Dragonwing™ IQ-9075 | 5.768 ms | 1 - 4 MB | NPU | LiteHRNet | ONNX | float | Qualcomm® Dragonwing™ IQ-X7181 | 5.636 ms | 5 - 5 MB | NPU | LiteHRNet | ONNX | float | Qualcomm® Dragonwing™ Q-8750 | 2.669 ms | 0 - 100 MB | NPU | LiteHRNet | QNN_DLC | float | Snapdragon® 8 Elite Gen 5 For Galaxy Mobile | 0.902 ms | 1 - 85 MB | NPU | LiteHRNet | QNN_DLC | float | Snapdragon® 8 Elite For Galaxy Mobile | 1.072 ms | 1 - 83 MB | NPU | LiteHRNet | QNN_DLC | float | Snapdragon® X2 Elite | 1.335 ms | 1 - 1 MB | NPU | LiteHRNet | QNN_DLC | float | Snapdragon® X Elite | 2.407 ms | 1 - 1 MB | NPU | LiteHRNet | QNN_DLC | float | Snapdragon® 8 Gen 3 Mobile | 1.403 ms | 0 - 99 MB | NPU | LiteHRNet | QNN_DLC | float | Snapdragon® 8 Gen 1 Mobile | 2.887 ms | 1 - 101 MB | NPU | LiteHRNet | QNN_DLC | float | Qualcomm® Dragonwing™ IQ-8275 | 2.171 ms | 1 - 4 MB | NPU | LiteHRNet | QNN_DLC | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 2.093 ms | 1 - 3 MB | NPU | LiteHRNet | QNN_DLC | float | Qualcomm® QCS8450 | 2.887 ms | 1 - 101 MB | NPU | LiteHRNet | QNN_DLC | float | Qualcomm® Dragonwing™ IQ-9075 | 2.478 ms | 3 - 5 MB | NPU | LiteHRNet | QNN_DLC | float | Qualcomm® Dragonwing™ IQ-X7181 | 2.407 ms | 1 - 1 MB | NPU | LiteHRNet | QNN_DLC | float | Qualcomm® Dragonwing™ Q-8750 | 1.072 ms | 1 - 83 MB | NPU | LiteHRNet | QNN_DLC | float | Qualcomm® SA8295P | 3.345 ms | 0 - 81 MB | NPU | LiteHRNet | TFLITE | float | Snapdragon® 8 Elite Gen 5 For Galaxy Mobile | 2.057 ms | 0 - 110 MB | NPU | LiteHRNet | TFLITE | float | Snapdragon® 8 Elite For Galaxy Mobile | 2.229 ms | 0 - 114 MB | NPU | LiteHRNet | TFLITE | float | Snapdragon® 8 Gen 3 Mobile | 2.672 ms | 0 - 146 MB | NPU | LiteHRNet | TFLITE | float | Snapdragon® 8 Gen 1 Mobile | 5.209 ms | 1 - 135 MB | NPU | LiteHRNet | TFLITE | float | Qualcomm® Dragonwing™ IQ-8275 | 4.204 ms | 1 - 12 MB | NPU | LiteHRNet | TFLITE | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 4.227 ms | 0 - 3 MB | NPU | LiteHRNet | TFLITE | float | Qualcomm® SA8775P | 5.278 ms | 1 - 112 MB | NPU | LiteHRNet | TFLITE | float | Qualcomm® SA8650P | 5.278 ms | 1 - 112 MB | NPU | LiteHRNet | TFLITE | float | Qualcomm® SA8255P | 5.278 ms | 1 - 112 MB | NPU | LiteHRNet | TFLITE | float | Qualcomm® QCS8450 | 5.209 ms | 1 - 135 MB | NPU | LiteHRNet | TFLITE | float | Qualcomm® Dragonwing™ IQ-9075 | 4.918 ms | 1 - 12 MB | NPU | LiteHRNet | TFLITE | float | Qualcomm® Dragonwing™ Q-8750 | 2.229 ms | 0 - 114 MB | NPU | LiteHRNet | TFLITE | float | Qualcomm® SA7255P | 8.598 ms | 1 - 111 MB | NPU | LiteHRNet | TFLITE | float | Qualcomm® SA8295P | 6.016 ms | 1 - 111 MB | NPU ## License * The license for the original implementation of LiteHRNet can be found [here](https://github.com/HRNet/Lite-HRNet/blob/hrnet/LICENSE). ## References * [Lite-HRNet: A Lightweight High-Resolution Network](https://arxiv.org/abs/2104.06403) * [Source Model Implementation](https://github.com/HRNet/Lite-HRNet) ## 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).