--- library_name: pytorch license: other tags: - bu_auto - android pipeline_tag: keypoint-detection --- ![](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/hrnet_pose/web-assets/model_demo.png) # HRNetPose: Optimized for Qualcomm Devices HRNet performs pose estimation in high-resolution representations. This is based on the implementation of HRNetPose found [here](https://github.com/leoxiaobin/deep-high-resolution-net.pytorch). 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/v0.58.0/src/qai_hub_models/models/hrnet_pose) 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 2.45, ONNX Runtime 1.25.0 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/hrnet_pose/releases/v0.58.0/hrnet_pose-onnx-float.zip) | ONNX | w8a16 | Universal | QAIRT 2.45, ONNX Runtime 1.25.0 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/hrnet_pose/releases/v0.58.0/hrnet_pose-onnx-w8a16.zip) | ONNX | w8a8 | Universal | QAIRT 2.45, ONNX Runtime 1.25.0 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/hrnet_pose/releases/v0.58.0/hrnet_pose-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/hrnet_pose/releases/v0.58.0/hrnet_pose-qnn_dlc-float.zip) | QNN_DLC | w8a16 | Universal | QAIRT 2.45 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/hrnet_pose/releases/v0.58.0/hrnet_pose-qnn_dlc-w8a16.zip) | QNN_DLC | w8a8 | Universal | QAIRT 2.45 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/hrnet_pose/releases/v0.58.0/hrnet_pose-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/hrnet_pose/releases/v0.58.0/hrnet_pose-tflite-float.zip) | TFLITE | w8a8 | Universal | QAIRT 2.45 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/hrnet_pose/releases/v0.58.0/hrnet_pose-tflite-w8a8.zip) For more device-specific assets and performance metrics, visit **[HRNetPose on Qualcomm® AI Hub](https://aihub.qualcomm.com/models/hrnet_pose)**. ### Option 2: Export with Custom Configurations Use the [Qualcomm® AI Hub Models](https://github.com/qualcomm/ai-hub-models/blob/v0.58.0/src/qai_hub_models/models/hrnet_pose) 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 [HRNetPose on GitHub](https://github.com/qualcomm/ai-hub-models/blob/v0.58.0/src/qai_hub_models/models/hrnet_pose) for usage instructions. ## Model Details **Model Type:** Model_use_case.pose_estimation **Model Stats:** - Model checkpoint: hrnet_posenet_FP32_state_dict - Input resolution: 256x192 - Number of parameters: 28.5M - Model size (float): 109 MB - Model size (w8a8): 28.1 MB ## Performance Summary | Model | Runtime | Precision | Chipset | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit |---|---|---|---|---|---|--- | HRNetPose | ONNX | float | Snapdragon® X2 Elite | 1.318 ms | 2 - 2 MB | NPU | HRNetPose | ONNX | float | Snapdragon® X Elite | 2.415 ms | 56 - 56 MB | NPU | HRNetPose | ONNX | float | Snapdragon® 8 Gen 3 Mobile | 1.855 ms | 0 - 141 MB | NPU | HRNetPose | ONNX | float | Snapdragon® 8 Gen 1 Mobile | 4.311 ms | 1 - 130 MB | NPU | HRNetPose | ONNX | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 2.441 ms | 0 - 58 MB | NPU | HRNetPose | ONNX | float | Qualcomm® QCS8450 | 4.311 ms | 1 - 130 MB | NPU | HRNetPose | ONNX | float | Snapdragon® 8 Elite Gen 5 Mobile | 1.245 ms | 1 - 87 MB | NPU | HRNetPose | ONNX | float | Snapdragon® 8 Elite Mobile | 1.447 ms | 0 - 86 MB | NPU | HRNetPose | ONNX | float | Qualcomm® Dragonwing™ Q-8750 | 1.447 ms | 0 - 86 MB | NPU | HRNetPose | ONNX | float | Qualcomm® Dragonwing™ IQ-X7181 | 2.415 ms | 56 - 56 MB | NPU | HRNetPose | ONNX | float | Qualcomm® Dragonwing™ IQ-9075 | 3.895 ms | 1 - 46 MB | NPU | HRNetPose | ONNX | w8a16 | Snapdragon® X2 Elite | 1.006 ms | 1 - 1 MB | NPU | HRNetPose | ONNX | w8a16 | Snapdragon® X Elite | 1.758 ms | 29 - 29 MB | NPU | HRNetPose | ONNX | w8a16 | Snapdragon® 8 Gen 3 Mobile | 1.23 ms | 0 - 200 MB | NPU | HRNetPose | ONNX | w8a16 | Snapdragon® 8 Gen 1 Mobile | 2.435 ms | 0 - 202 MB | NPU | HRNetPose | ONNX | w8a16 | Qualcomm® Dragonwing™ QCS6490 | 8.199 ms | 0 - 3 MB | NPU | HRNetPose | ONNX | w8a16 | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 1.717 ms | 0 - 35 MB | NPU | HRNetPose | ONNX | w8a16 | Qualcomm® QCS8450 | 2.435 ms | 0 - 202 MB | NPU | HRNetPose | ONNX | w8a16 | Snapdragon® 8 Elite Mobile | 0.97 ms | 0 - 134 MB | NPU | HRNetPose | ONNX | w8a16 | Qualcomm® Dragonwing™ IQ-9075 | 1.897 ms | 0 - 3 MB | NPU | HRNetPose | ONNX | w8a16 | Snapdragon® 7 Gen 4 Mobile | 2.322 ms | 0 - 137 MB | NPU | HRNetPose | ONNX | w8a16 | Snapdragon® 8 Elite Gen 5 Mobile | 0.816 ms | 0 - 136 MB | NPU | HRNetPose | ONNX | w8a16 | Qualcomm® Dragonwing™ Q-6690 | 24.344 ms | 0 - 251 MB | NPU | HRNetPose | ONNX | w8a16 | Qualcomm® Dragonwing™ Q-7790 | 2.322 ms | 0 - 137 MB | NPU | HRNetPose | ONNX | w8a16 | Qualcomm® Dragonwing™ Q-8750 | 0.97 ms | 0 - 134 MB | NPU | HRNetPose | ONNX | w8a16 | Qualcomm® Dragonwing™ IQ-X7181 | 1.758 ms | 29 - 29 MB | NPU | HRNetPose | ONNX | w8a8 | Snapdragon® X2 Elite | 0.57 ms | 1 - 1 MB | NPU | HRNetPose | ONNX | w8a8 | Snapdragon® X Elite | 1.121 ms | 28 - 28 MB | NPU | HRNetPose | ONNX | w8a8 | Snapdragon® 8 Gen 3 Mobile | 0.852 ms | 0 - 186 MB | NPU | HRNetPose | ONNX | w8a8 | Snapdragon® 8 Gen 1 Mobile | 1.748 ms | 0 - 187 MB | NPU | HRNetPose | ONNX | w8a8 | Qualcomm® Dragonwing™ QCS6490 | 5.035 ms | 0 - 3 MB | NPU | HRNetPose | ONNX | w8a8 | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 1.142 ms | 0 - 30 MB | NPU | HRNetPose | ONNX | w8a8 | Qualcomm® QCS8450 | 1.748 ms | 0 - 187 MB | NPU | HRNetPose | ONNX | w8a8 | Snapdragon® 8 Elite Mobile | 0.723 ms | 0 - 124 MB | NPU | HRNetPose | ONNX | w8a8 | Snapdragon® 7 Gen 4 Mobile | 1.497 ms | 0 - 125 MB | NPU | HRNetPose | ONNX | w8a8 | Snapdragon® 8 Elite Gen 5 Mobile | 0.636 ms | 0 - 126 MB | NPU | HRNetPose | ONNX | w8a8 | Qualcomm® Dragonwing™ Q-6690 | 10.426 ms | 0 - 239 MB | NPU | HRNetPose | ONNX | w8a8 | Qualcomm® Dragonwing™ Q-7790 | 1.497 ms | 0 - 125 MB | NPU | HRNetPose | ONNX | w8a8 | Qualcomm® Dragonwing™ Q-8750 | 0.723 ms | 0 - 124 MB | NPU | HRNetPose | ONNX | w8a8 | Qualcomm® Dragonwing™ IQ-X7181 | 1.121 ms | 28 - 28 MB | NPU | HRNetPose | ONNX | w8a8 | Qualcomm® Dragonwing™ IQ-9075 | 1.302 ms | 0 - 45 MB | NPU | HRNetPose | QNN_DLC | float | Snapdragon® X2 Elite | 1.61 ms | 1 - 1 MB | NPU | HRNetPose | QNN_DLC | float | Snapdragon® X Elite | 3.005 ms | 1 - 1 MB | NPU | HRNetPose | QNN_DLC | float | Snapdragon® 8 Gen 3 Mobile | 1.989 ms | 0 - 114 MB | NPU | HRNetPose | QNN_DLC | float | Snapdragon® 8 Gen 1 Mobile | 5.037 ms | 0 - 104 MB | NPU | HRNetPose | QNN_DLC | float | Qualcomm® QCS8275 | 14.171 ms | 1 - 73 MB | NPU | HRNetPose | QNN_DLC | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 2.773 ms | 1 - 172 MB | NPU | HRNetPose | QNN_DLC | float | Qualcomm® SA8775P | 4.344 ms | 1 - 75 MB | NPU | HRNetPose | QNN_DLC | float | Qualcomm® SA8650P | 4.344 ms | 1 - 75 MB | NPU | HRNetPose | QNN_DLC | float | Qualcomm® SA8255P | 4.344 ms | 1 - 75 MB | NPU | HRNetPose | QNN_DLC | float | Qualcomm® QCS8450 | 5.037 ms | 0 - 104 MB | NPU | HRNetPose | QNN_DLC | float | Snapdragon® 8 Elite Gen 5 Mobile | 1.292 ms | 0 - 76 MB | NPU | HRNetPose | QNN_DLC | float | Snapdragon® 8 Elite Mobile | 1.563 ms | 0 - 74 MB | NPU | HRNetPose | QNN_DLC | float | Qualcomm® SA7255P | 14.171 ms | 1 - 73 MB | NPU | HRNetPose | QNN_DLC | float | Qualcomm® SA8295P | 4.601 ms | 1 - 66 MB | NPU | HRNetPose | QNN_DLC | float | Qualcomm® Dragonwing™ Q-8750 | 1.563 ms | 0 - 74 MB | NPU | HRNetPose | QNN_DLC | float | Qualcomm® Dragonwing™ IQ-X7181 | 3.005 ms | 1 - 1 MB | NPU | HRNetPose | QNN_DLC | float | Qualcomm® Dragonwing™ IQ-9075 | 4.118 ms | 1 - 3 MB | NPU | HRNetPose | QNN_DLC | w8a16 | Snapdragon® X2 Elite | 1.324 ms | 0 - 0 MB | NPU | HRNetPose | QNN_DLC | w8a16 | Snapdragon® X Elite | 2.124 ms | 0 - 0 MB | NPU | HRNetPose | QNN_DLC | w8a16 | Snapdragon® 8 Gen 3 Mobile | 1.388 ms | 0 - 148 MB | NPU | HRNetPose | QNN_DLC | w8a16 | Snapdragon® 8 Gen 1 Mobile | 2.733 ms | 0 - 149 MB | NPU | HRNetPose | QNN_DLC | w8a16 | Qualcomm® Dragonwing™ QCS6490 | 8.368 ms | 2 - 4 MB | NPU | HRNetPose | QNN_DLC | w8a16 | Qualcomm® QCS8275 | 5.163 ms | 0 - 99 MB | NPU | HRNetPose | QNN_DLC | w8a16 | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 1.908 ms | 0 - 2 MB | NPU | HRNetPose | QNN_DLC | w8a16 | Qualcomm® SA8775P | 2.267 ms | 0 - 102 MB | NPU | HRNetPose | QNN_DLC | w8a16 | Qualcomm® SA8650P | 2.267 ms | 0 - 102 MB | NPU | HRNetPose | QNN_DLC | w8a16 | Qualcomm® SA8255P | 2.267 ms | 0 - 102 MB | NPU | HRNetPose | QNN_DLC | w8a16 | Qualcomm® QCS8450 | 2.733 ms | 0 - 149 MB | NPU | HRNetPose | QNN_DLC | w8a16 | Snapdragon® 8 Elite Mobile | 0.999 ms | 0 - 102 MB | NPU | HRNetPose | QNN_DLC | w8a16 | Qualcomm® SA7255P | 5.163 ms | 0 - 99 MB | NPU | HRNetPose | QNN_DLC | w8a16 | Qualcomm® SA8295P | 3.061 ms | 0 - 98 MB | NPU | HRNetPose | QNN_DLC | w8a16 | Qualcomm® Dragonwing™ IQ-9075 | 2.184 ms | 2 - 4 MB | NPU | HRNetPose | QNN_DLC | w8a16 | Snapdragon® 7 Gen 4 Mobile | 2.549 ms | 0 - 100 MB | NPU | HRNetPose | QNN_DLC | w8a16 | Snapdragon® 8 Elite Gen 5 Mobile | 0.836 ms | 0 - 101 MB | NPU | HRNetPose | QNN_DLC | w8a16 | Qualcomm® Dragonwing™ Q-6690 | 19.639 ms | 0 - 217 MB | NPU | HRNetPose | QNN_DLC | w8a16 | Qualcomm® Dragonwing™ Q-7790 | 2.549 ms | 0 - 100 MB | NPU | HRNetPose | QNN_DLC | w8a16 | Qualcomm® Dragonwing™ Q-8750 | 0.999 ms | 0 - 102 MB | NPU | HRNetPose | QNN_DLC | w8a16 | Qualcomm® Dragonwing™ IQ-X7181 | 2.124 ms | 0 - 0 MB | NPU | HRNetPose | QNN_DLC | w8a8 | Snapdragon® X2 Elite | 0.696 ms | 0 - 0 MB | NPU | HRNetPose | QNN_DLC | w8a8 | Snapdragon® X Elite | 1.239 ms | 0 - 0 MB | NPU | HRNetPose | QNN_DLC | w8a8 | Snapdragon® 8 Gen 3 Mobile | 0.805 ms | 0 - 134 MB | NPU | HRNetPose | QNN_DLC | w8a8 | Snapdragon® 8 Gen 1 Mobile | 1.738 ms | 0 - 136 MB | NPU | HRNetPose | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ QCS6490 | 5.39 ms | 0 - 2 MB | NPU | HRNetPose | QNN_DLC | w8a8 | Qualcomm® QCS8275 | 2.822 ms | 0 - 90 MB | NPU | HRNetPose | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 1.118 ms | 0 - 2 MB | NPU | HRNetPose | QNN_DLC | w8a8 | Qualcomm® SA8775P | 1.489 ms | 0 - 92 MB | NPU | HRNetPose | QNN_DLC | w8a8 | Qualcomm® SA8650P | 1.489 ms | 0 - 92 MB | NPU | HRNetPose | QNN_DLC | w8a8 | Qualcomm® SA8255P | 1.489 ms | 0 - 92 MB | NPU | HRNetPose | QNN_DLC | w8a8 | Qualcomm® QCS8450 | 1.738 ms | 0 - 136 MB | NPU | HRNetPose | QNN_DLC | w8a8 | Snapdragon® 8 Elite Mobile | 0.635 ms | 0 - 94 MB | NPU | HRNetPose | QNN_DLC | w8a8 | Snapdragon® 7 Gen 4 Mobile | 1.504 ms | 0 - 94 MB | NPU | HRNetPose | QNN_DLC | w8a8 | Qualcomm® SA7255P | 2.822 ms | 0 - 90 MB | NPU | HRNetPose | QNN_DLC | w8a8 | Qualcomm® SA8295P | 1.958 ms | 0 - 88 MB | NPU | HRNetPose | QNN_DLC | w8a8 | Snapdragon® 8 Elite Gen 5 Mobile | 0.565 ms | 0 - 95 MB | NPU | HRNetPose | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ Q-6690 | 10.462 ms | 0 - 213 MB | NPU | HRNetPose | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ Q-7790 | 1.504 ms | 0 - 94 MB | NPU | HRNetPose | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ Q-8750 | 0.635 ms | 0 - 94 MB | NPU | HRNetPose | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ IQ-X7181 | 1.239 ms | 0 - 0 MB | NPU | HRNetPose | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ IQ-9075 | 1.273 ms | 0 - 2 MB | NPU | HRNetPose | TFLITE | float | Snapdragon® 8 Gen 3 Mobile | 1.987 ms | 0 - 182 MB | NPU | HRNetPose | TFLITE | float | Snapdragon® 8 Gen 1 Mobile | 5.004 ms | 0 - 174 MB | NPU | HRNetPose | TFLITE | float | Qualcomm® QCS8275 | 14.125 ms | 0 - 112 MB | NPU | HRNetPose | TFLITE | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 2.727 ms | 0 - 2 MB | NPU | HRNetPose | TFLITE | float | Qualcomm® SA8775P | 4.324 ms | 0 - 114 MB | NPU | HRNetPose | TFLITE | float | Qualcomm® SA8650P | 4.324 ms | 0 - 114 MB | NPU | HRNetPose | TFLITE | float | Qualcomm® SA8255P | 4.324 ms | 0 - 114 MB | NPU | HRNetPose | TFLITE | float | Qualcomm® QCS8450 | 5.004 ms | 0 - 174 MB | NPU | HRNetPose | TFLITE | float | Snapdragon® 8 Elite Gen 5 Mobile | 1.302 ms | 0 - 115 MB | NPU | HRNetPose | TFLITE | float | Snapdragon® 8 Elite Mobile | 1.55 ms | 0 - 114 MB | NPU | HRNetPose | TFLITE | float | Qualcomm® SA7255P | 14.125 ms | 0 - 112 MB | NPU | HRNetPose | TFLITE | float | Qualcomm® SA8295P | 4.633 ms | 0 - 103 MB | NPU | HRNetPose | TFLITE | float | Qualcomm® Dragonwing™ Q-8750 | 1.55 ms | 0 - 114 MB | NPU | HRNetPose | TFLITE | float | Qualcomm® Dragonwing™ IQ-9075 | 4.128 ms | 0 - 58 MB | NPU | HRNetPose | TFLITE | w8a8 | Snapdragon® 8 Gen 3 Mobile | 0.728 ms | 3 - 144 MB | NPU | HRNetPose | TFLITE | w8a8 | Snapdragon® 8 Gen 1 Mobile | 1.515 ms | 0 - 142 MB | NPU | HRNetPose | TFLITE | w8a8 | Qualcomm® Dragonwing™ QCS6490 | 4.972 ms | 0 - 30 MB | NPU | HRNetPose | TFLITE | w8a8 | Qualcomm® QCS8275 | 2.59 ms | 0 - 88 MB | NPU | HRNetPose | TFLITE | w8a8 | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 0.947 ms | 0 - 3 MB | NPU | HRNetPose | TFLITE | w8a8 | Qualcomm® SA8775P | 1.339 ms | 0 - 89 MB | NPU | HRNetPose | TFLITE | w8a8 | Qualcomm® SA8650P | 1.339 ms | 0 - 89 MB | NPU | HRNetPose | TFLITE | w8a8 | Qualcomm® SA8255P | 1.339 ms | 0 - 89 MB | NPU | HRNetPose | TFLITE | w8a8 | Qualcomm® QCS8450 | 1.515 ms | 0 - 142 MB | NPU | HRNetPose | TFLITE | w8a8 | Snapdragon® 8 Elite Mobile | 0.579 ms | 0 - 88 MB | NPU | HRNetPose | TFLITE | w8a8 | Snapdragon® 7 Gen 4 Mobile | 1.355 ms | 0 - 85 MB | NPU | HRNetPose | TFLITE | w8a8 | Qualcomm® SA7255P | 2.59 ms | 0 - 88 MB | NPU | HRNetPose | TFLITE | w8a8 | Qualcomm® SA8295P | 1.767 ms | 0 - 85 MB | NPU | HRNetPose | TFLITE | w8a8 | Snapdragon® 8 Elite Gen 5 Mobile | 0.515 ms | 0 - 93 MB | NPU | HRNetPose | TFLITE | w8a8 | Qualcomm® Dragonwing™ Q-6690 | 9.633 ms | 0 - 207 MB | NPU | HRNetPose | TFLITE | w8a8 | Qualcomm® Dragonwing™ Q-7790 | 1.355 ms | 0 - 85 MB | NPU | HRNetPose | TFLITE | w8a8 | Qualcomm® Dragonwing™ Q-8750 | 0.579 ms | 0 - 88 MB | NPU | HRNetPose | TFLITE | w8a8 | Qualcomm® Dragonwing™ IQ-9075 | 1.08 ms | 0 - 30 MB | NPU ## License * The license for the original implementation of HRNetPose can be found [here](https://github.com/leoxiaobin/deep-high-resolution-net.pytorch/blob/master/LICENSE). ## References * [Deep High-Resolution Representation Learning for Human Pose Estimation](https://arxiv.org/abs/1902.09212) * [Source Model Implementation](https://github.com/leoxiaobin/deep-high-resolution-net.pytorch) ## 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).