v0.49.1
Browse filesSee https://github.com/qualcomm/ai-hub-models/releases/v0.49.1 for changelog.
README.md
CHANGED
|
@@ -14,7 +14,7 @@ pipeline_tag: keypoint-detection
|
|
| 14 |
LiteHRNet is a machine learning model that detects human pose and returns a location and confidence for each of 17 joints.
|
| 15 |
|
| 16 |
This is based on the implementation of LiteHRNet found [here](https://github.com/HRNet/Lite-HRNet).
|
| 17 |
-
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/
|
| 18 |
|
| 19 |
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.
|
| 20 |
|
|
@@ -27,23 +27,23 @@ Below are pre-exported model assets ready for deployment.
|
|
| 27 |
|
| 28 |
| Runtime | Precision | Chipset | SDK Versions | Download |
|
| 29 |
|---|---|---|---|---|
|
| 30 |
-
| ONNX | float | Universal | QAIRT 2.42, ONNX Runtime 1.24.1 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/litehrnet/releases/v0.
|
| 31 |
-
| QNN_DLC | float | Universal | QAIRT 2.43 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/litehrnet/releases/v0.
|
| 32 |
-
| TFLITE | float | Universal | QAIRT 2.43, TFLite 2.17.0 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/litehrnet/releases/v0.
|
| 33 |
|
| 34 |
For more device-specific assets and performance metrics, visit **[LiteHRNet on Qualcomm® AI Hub](https://aihub.qualcomm.com/models/litehrnet)**.
|
| 35 |
|
| 36 |
|
| 37 |
### Option 2: Export with Custom Configurations
|
| 38 |
|
| 39 |
-
Use the [Qualcomm® AI Hub Models](https://github.com/qualcomm/ai-hub-models/
|
| 40 |
- Custom weights (e.g., fine-tuned checkpoints)
|
| 41 |
- Custom input shapes
|
| 42 |
- Target device and runtime configurations
|
| 43 |
|
| 44 |
This option is ideal if you need to customize the model beyond the default configuration provided here.
|
| 45 |
|
| 46 |
-
See our repository for [LiteHRNet on GitHub](https://github.com/qualcomm/ai-hub-models/
|
| 47 |
|
| 48 |
## Model Details
|
| 49 |
|
|
@@ -57,34 +57,35 @@ See our repository for [LiteHRNet on GitHub](https://github.com/qualcomm/ai-hub-
|
|
| 57 |
## Performance Summary
|
| 58 |
| Model | Runtime | Precision | Chipset | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit
|
| 59 |
|---|---|---|---|---|---|---
|
| 60 |
-
| LiteHRNet | ONNX | float | Snapdragon®
|
| 61 |
-
| LiteHRNet | ONNX | float | Snapdragon®
|
| 62 |
-
| LiteHRNet | ONNX | float | Snapdragon®
|
| 63 |
-
| LiteHRNet | ONNX | float |
|
| 64 |
-
| LiteHRNet | ONNX | float |
|
| 65 |
-
| LiteHRNet | ONNX | float |
|
| 66 |
-
| LiteHRNet |
|
| 67 |
-
| LiteHRNet | QNN_DLC | float | Snapdragon®
|
| 68 |
-
| LiteHRNet | QNN_DLC | float | Snapdragon®
|
| 69 |
-
| LiteHRNet | QNN_DLC | float |
|
| 70 |
-
| LiteHRNet | QNN_DLC | float |
|
| 71 |
-
| LiteHRNet | QNN_DLC | float | Qualcomm®
|
| 72 |
-
| LiteHRNet | QNN_DLC | float | Qualcomm®
|
| 73 |
-
| LiteHRNet | QNN_DLC | float | Qualcomm®
|
| 74 |
-
| LiteHRNet | QNN_DLC | float | Qualcomm®
|
| 75 |
-
| LiteHRNet | QNN_DLC | float | Qualcomm®
|
| 76 |
-
| LiteHRNet | QNN_DLC | float |
|
| 77 |
-
| LiteHRNet | QNN_DLC | float |
|
| 78 |
-
| LiteHRNet |
|
| 79 |
-
| LiteHRNet | TFLITE | float |
|
| 80 |
-
| LiteHRNet | TFLITE | float |
|
| 81 |
-
| LiteHRNet | TFLITE | float | Qualcomm®
|
| 82 |
-
| LiteHRNet | TFLITE | float | Qualcomm®
|
| 83 |
-
| LiteHRNet | TFLITE | float | Qualcomm®
|
| 84 |
-
| LiteHRNet | TFLITE | float | Qualcomm®
|
| 85 |
-
| LiteHRNet | TFLITE | float | Qualcomm®
|
| 86 |
-
| LiteHRNet | TFLITE | float |
|
| 87 |
-
| LiteHRNet | TFLITE | float |
|
|
|
|
| 88 |
|
| 89 |
## License
|
| 90 |
* The license for the original implementation of LiteHRNet can be found
|
|
|
|
| 14 |
LiteHRNet is a machine learning model that detects human pose and returns a location and confidence for each of 17 joints.
|
| 15 |
|
| 16 |
This is based on the implementation of LiteHRNet found [here](https://github.com/HRNet/Lite-HRNet).
|
| 17 |
+
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/tree/v0.49.1/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).
|
| 18 |
|
| 19 |
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.
|
| 20 |
|
|
|
|
| 27 |
|
| 28 |
| Runtime | Precision | Chipset | SDK Versions | Download |
|
| 29 |
|---|---|---|---|---|
|
| 30 |
+
| ONNX | float | Universal | QAIRT 2.42, ONNX Runtime 1.24.1 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/litehrnet/releases/v0.49.1/litehrnet-onnx-float.zip)
|
| 31 |
+
| QNN_DLC | float | Universal | QAIRT 2.43 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/litehrnet/releases/v0.49.1/litehrnet-qnn_dlc-float.zip)
|
| 32 |
+
| TFLITE | float | Universal | QAIRT 2.43, TFLite 2.17.0 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/litehrnet/releases/v0.49.1/litehrnet-tflite-float.zip)
|
| 33 |
|
| 34 |
For more device-specific assets and performance metrics, visit **[LiteHRNet on Qualcomm® AI Hub](https://aihub.qualcomm.com/models/litehrnet)**.
|
| 35 |
|
| 36 |
|
| 37 |
### Option 2: Export with Custom Configurations
|
| 38 |
|
| 39 |
+
Use the [Qualcomm® AI Hub Models](https://github.com/qualcomm/ai-hub-models/tree/v0.49.1/qai_hub_models/models/litehrnet) Python library to compile and export the model with your own:
|
| 40 |
- Custom weights (e.g., fine-tuned checkpoints)
|
| 41 |
- Custom input shapes
|
| 42 |
- Target device and runtime configurations
|
| 43 |
|
| 44 |
This option is ideal if you need to customize the model beyond the default configuration provided here.
|
| 45 |
|
| 46 |
+
See our repository for [LiteHRNet on GitHub](https://github.com/qualcomm/ai-hub-models/tree/v0.49.1/qai_hub_models/models/litehrnet) for usage instructions.
|
| 47 |
|
| 48 |
## Model Details
|
| 49 |
|
|
|
|
| 57 |
## Performance Summary
|
| 58 |
| Model | Runtime | Precision | Chipset | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit
|
| 59 |
|---|---|---|---|---|---|---
|
| 60 |
+
| LiteHRNet | ONNX | float | Snapdragon® 8 Elite Gen 5 Mobile | 2.738 ms | 1 - 98 MB | NPU
|
| 61 |
+
| LiteHRNet | ONNX | float | Snapdragon® X2 Elite | 2.852 ms | 5 - 5 MB | NPU
|
| 62 |
+
| LiteHRNet | ONNX | float | Snapdragon® X Elite | 5.733 ms | 5 - 5 MB | NPU
|
| 63 |
+
| LiteHRNet | ONNX | float | Snapdragon® 8 Gen 3 Mobile | 3.145 ms | 0 - 129 MB | NPU
|
| 64 |
+
| LiteHRNet | ONNX | float | Qualcomm® QCS8550 (Proxy) | 5.198 ms | 0 - 12 MB | NPU
|
| 65 |
+
| LiteHRNet | ONNX | float | Qualcomm® QCS9075 | 5.904 ms | 1 - 4 MB | NPU
|
| 66 |
+
| LiteHRNet | ONNX | float | Snapdragon® 8 Elite For Galaxy Mobile | 2.848 ms | 0 - 100 MB | NPU
|
| 67 |
+
| LiteHRNet | QNN_DLC | float | Snapdragon® 8 Elite Gen 5 Mobile | 0.838 ms | 1 - 84 MB | NPU
|
| 68 |
+
| LiteHRNet | QNN_DLC | float | Snapdragon® X2 Elite | 1.228 ms | 1 - 1 MB | NPU
|
| 69 |
+
| LiteHRNet | QNN_DLC | float | Snapdragon® X Elite | 2.392 ms | 1 - 1 MB | NPU
|
| 70 |
+
| LiteHRNet | QNN_DLC | float | Snapdragon® 8 Gen 3 Mobile | 1.367 ms | 1 - 111 MB | NPU
|
| 71 |
+
| LiteHRNet | QNN_DLC | float | Qualcomm® QCS8275 (Proxy) | 4.937 ms | 1 - 81 MB | NPU
|
| 72 |
+
| LiteHRNet | QNN_DLC | float | Qualcomm® QCS8550 (Proxy) | 2.07 ms | 1 - 2 MB | NPU
|
| 73 |
+
| LiteHRNet | QNN_DLC | float | Qualcomm® SA8775P | 10.887 ms | 1 - 80 MB | NPU
|
| 74 |
+
| LiteHRNet | QNN_DLC | float | Qualcomm® QCS9075 | 2.512 ms | 1 - 3 MB | NPU
|
| 75 |
+
| LiteHRNet | QNN_DLC | float | Qualcomm® QCS8450 (Proxy) | 2.806 ms | 0 - 109 MB | NPU
|
| 76 |
+
| LiteHRNet | QNN_DLC | float | Qualcomm® SA7255P | 4.937 ms | 1 - 81 MB | NPU
|
| 77 |
+
| LiteHRNet | QNN_DLC | float | Qualcomm® SA8295P | 3.472 ms | 0 - 83 MB | NPU
|
| 78 |
+
| LiteHRNet | QNN_DLC | float | Snapdragon® 8 Elite For Galaxy Mobile | 1.016 ms | 0 - 84 MB | NPU
|
| 79 |
+
| LiteHRNet | TFLITE | float | Snapdragon® 8 Elite Gen 5 Mobile | 2.015 ms | 0 - 120 MB | NPU
|
| 80 |
+
| LiteHRNet | TFLITE | float | Snapdragon® 8 Gen 3 Mobile | 2.715 ms | 0 - 150 MB | NPU
|
| 81 |
+
| LiteHRNet | TFLITE | float | Qualcomm® QCS8275 (Proxy) | 8.753 ms | 0 - 114 MB | NPU
|
| 82 |
+
| LiteHRNet | TFLITE | float | Qualcomm® QCS8550 (Proxy) | 4.212 ms | 0 - 2 MB | NPU
|
| 83 |
+
| LiteHRNet | TFLITE | float | Qualcomm® SA8775P | 5.16 ms | 0 - 115 MB | NPU
|
| 84 |
+
| LiteHRNet | TFLITE | float | Qualcomm® QCS9075 | 5.075 ms | 0 - 10 MB | NPU
|
| 85 |
+
| LiteHRNet | TFLITE | float | Qualcomm® QCS8450 (Proxy) | 5.327 ms | 0 - 136 MB | NPU
|
| 86 |
+
| LiteHRNet | TFLITE | float | Qualcomm® SA7255P | 8.753 ms | 0 - 114 MB | NPU
|
| 87 |
+
| LiteHRNet | TFLITE | float | Qualcomm® SA8295P | 6.292 ms | 0 - 114 MB | NPU
|
| 88 |
+
| LiteHRNet | TFLITE | float | Snapdragon® 8 Elite For Galaxy Mobile | 2.201 ms | 0 - 119 MB | NPU
|
| 89 |
|
| 90 |
## License
|
| 91 |
* The license for the original implementation of LiteHRNet can be found
|