YamNet: Optimized for Qualcomm Devices
An audio event classifier trained on the AudioSet dataset to predict audio events from the AudioSet ontology employing the Mobilenet_v1 depthwise-separable convolution architecture.
This is based on the implementation of YamNet found here. This repository contains pre-exported model files optimized for Qualcomm® devices. You can use the Qualcomm® AI Hub Models library to export with custom configurations. More details on model performance across various devices, can be found here.
Qualcomm AI Hub Models uses Qualcomm AI Hub Workbench to compile, profile, and evaluate this model. Sign up 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 1.27.1 | Download |
| ONNX | w8a16 | Universal | QAIRT 2.45, ONNX Runtime 1.27.1 | Download |
| ONNX | w8a8 | Universal | QAIRT 2.45, ONNX Runtime 1.27.1 | Download |
| QNN_DLC | float | Universal | QAIRT 2.45 | Download |
| QNN_DLC | w8a16 | Universal | QAIRT 2.45 | Download |
| QNN_DLC | w8a8 | Universal | QAIRT 2.45 | Download |
| TFLITE | float | Universal | QAIRT 2.45 | Download |
| TFLITE | w8a8 | Universal | QAIRT 2.45 | Download |
For more device-specific assets and performance metrics, visit YamNet on Qualcomm® AI Hub.
Option 2: Export with Custom Configurations
Use the Qualcomm® AI Hub Models 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 YamNet on GitHub for usage instructions.
Model Details
Model Type: Model_use_case.audio_classification
Model Stats:
- Input resolution: 1x1x96x64
- Model checkpoint: yamnet.pth
- Model size (float): 14.2 MB
- Number of parameters: 3.73M
Performance Summary
| Model | Runtime | Precision | Chipset | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit |
|---|---|---|---|---|---|---|
| YamNet | ONNX | float | Snapdragon® X2 Elite | 0.092 ms | 1 - 1 MB | NPU |
| YamNet | ONNX | float | Snapdragon® X Elite | 0.166 ms | 8 - 8 MB | NPU |
| YamNet | ONNX | float | Snapdragon® 8 Gen 3 Mobile | 0.176 ms | 0 - 38 MB | NPU |
| YamNet | ONNX | float | Snapdragon® 8 Gen 1 Mobile | 0.362 ms | 0 - 42 MB | NPU |
| YamNet | ONNX | float | Qualcomm® Dragonwing™ IQ-8275 | 0.332 ms | 0 - 3 MB | NPU |
| YamNet | ONNX | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 0.221 ms | 0 - 9 MB | NPU |
| YamNet | ONNX | float | Qualcomm® QCS8450 | 0.362 ms | 0 - 42 MB | NPU |
| YamNet | ONNX | float | Qualcomm® Dragonwing™ IQ-9075 | 0.309 ms | 0 - 3 MB | NPU |
| YamNet | ONNX | float | Qualcomm® Dragonwing™ IQ-X7181 | 0.166 ms | 8 - 8 MB | NPU |
| YamNet | ONNX | float | Qualcomm® Dragonwing™ Q-8750 | 0.16 ms | 0 - 25 MB | NPU |
| YamNet | ONNX | float | Snapdragon® 8 Elite Mobile | 0.16 ms | 0 - 25 MB | NPU |
| YamNet | ONNX | float | Snapdragon® 8 Elite Gen 5 Mobile | 0.153 ms | 0 - 22 MB | NPU |
| YamNet | ONNX | w8a16 | Snapdragon® X2 Elite | 0.061 ms | 1 - 1 MB | NPU |
| YamNet | ONNX | w8a16 | Snapdragon® X Elite | 0.096 ms | 0 - 0 MB | NPU |
| YamNet | ONNX | w8a16 | Snapdragon® 8 Gen 3 Mobile | 0.108 ms | 0 - 39 MB | NPU |
| YamNet | ONNX | w8a16 | Snapdragon® 8 Gen 1 Mobile | 0.225 ms | 0 - 44 MB | NPU |
| YamNet | ONNX | w8a16 | Qualcomm® Dragonwing™ QCS6490 | 0.612 ms | 0 - 3 MB | NPU |
| YamNet | ONNX | w8a16 | Qualcomm® Dragonwing™ IQ-8275 | 0.339 ms | 0 - 4 MB | NPU |
| YamNet | ONNX | w8a16 | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 0.133 ms | 0 - 6 MB | NPU |
| YamNet | ONNX | w8a16 | Qualcomm® QCS8450 | 0.225 ms | 0 - 44 MB | NPU |
| YamNet | ONNX | w8a16 | Qualcomm® Dragonwing™ IQ-9075 | 0.226 ms | 0 - 3 MB | NPU |
| YamNet | ONNX | w8a16 | Qualcomm® Dragonwing™ IQ-X7181 | 0.096 ms | 0 - 0 MB | NPU |
| YamNet | ONNX | w8a16 | Qualcomm® Dragonwing™ Q-6690 | 0.502 ms | 0 - 25 MB | NPU |
| YamNet | ONNX | w8a16 | Qualcomm® Dragonwing™ Q-7790 | 0.168 ms | 0 - 25 MB | NPU |
| YamNet | ONNX | w8a16 | Qualcomm® Dragonwing™ Q-8750 | 0.1 ms | 0 - 30 MB | NPU |
| YamNet | ONNX | w8a16 | Snapdragon® 8 Elite Mobile | 0.1 ms | 0 - 30 MB | NPU |
| YamNet | ONNX | w8a16 | Snapdragon® 8 Elite Gen 5 Mobile | 0.096 ms | 0 - 26 MB | NPU |
| YamNet | ONNX | w8a16 | Snapdragon® 7 Gen 4 Mobile | 0.168 ms | 0 - 25 MB | NPU |
| YamNet | ONNX | w8a8 | Snapdragon® X2 Elite | 0.058 ms | 1 - 1 MB | NPU |
| YamNet | ONNX | w8a8 | Snapdragon® X Elite | 0.095 ms | 0 - 0 MB | NPU |
| YamNet | ONNX | w8a8 | Snapdragon® 8 Gen 3 Mobile | 0.107 ms | 0 - 39 MB | NPU |
| YamNet | ONNX | w8a8 | Snapdragon® 8 Gen 1 Mobile | 0.213 ms | 0 - 45 MB | NPU |
| YamNet | ONNX | w8a8 | Qualcomm® Dragonwing™ QCS6490 | 0.723 ms | 0 - 3 MB | NPU |
| YamNet | ONNX | w8a8 | Qualcomm® Dragonwing™ IQ-8275 | 0.317 ms | 0 - 3 MB | NPU |
| YamNet | ONNX | w8a8 | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 0.132 ms | 0 - 6 MB | NPU |
| YamNet | ONNX | w8a8 | Qualcomm® QCS8450 | 0.213 ms | 0 - 45 MB | NPU |
| YamNet | ONNX | w8a8 | Qualcomm® Dragonwing™ IQ-9075 | 0.219 ms | 0 - 3 MB | NPU |
| YamNet | ONNX | w8a8 | Qualcomm® Dragonwing™ IQ-X7181 | 0.095 ms | 0 - 0 MB | NPU |
| YamNet | ONNX | w8a8 | Qualcomm® Dragonwing™ Q-6690 | 0.44 ms | 0 - 25 MB | NPU |
| YamNet | ONNX | w8a8 | Qualcomm® Dragonwing™ Q-7790 | 0.165 ms | 0 - 25 MB | NPU |
| YamNet | ONNX | w8a8 | Qualcomm® Dragonwing™ Q-8750 | 0.097 ms | 0 - 24 MB | NPU |
| YamNet | ONNX | w8a8 | Snapdragon® 8 Elite Mobile | 0.097 ms | 0 - 24 MB | NPU |
| YamNet | ONNX | w8a8 | Snapdragon® 8 Elite Gen 5 Mobile | 0.099 ms | 0 - 24 MB | NPU |
| YamNet | ONNX | w8a8 | Snapdragon® 7 Gen 4 Mobile | 0.165 ms | 0 - 25 MB | NPU |
| YamNet | QNN_DLC | float | Snapdragon® X2 Elite | 0.195 ms | 0 - 0 MB | NPU |
| YamNet | QNN_DLC | float | Snapdragon® X Elite | 0.278 ms | 0 - 0 MB | NPU |
| YamNet | QNN_DLC | float | Snapdragon® 8 Gen 3 Mobile | 0.175 ms | 0 - 36 MB | NPU |
| YamNet | QNN_DLC | float | Snapdragon® 8 Gen 1 Mobile | 0.35 ms | 0 - 43 MB | NPU |
| YamNet | QNN_DLC | float | Qualcomm® Dragonwing™ IQ-8275 | 0.328 ms | 0 - 2 MB | NPU |
| YamNet | QNN_DLC | float | Qualcomm® Dragonwing™ IQ-8275 | 0.678 ms | 0 - 21 MB | NPU |
| YamNet | QNN_DLC | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 0.214 ms | 0 - 58 MB | NPU |
| YamNet | QNN_DLC | float | Qualcomm® SA8775P | 0.36 ms | 0 - 22 MB | NPU |
| YamNet | QNN_DLC | float | Qualcomm® SA8650P | 0.36 ms | 0 - 22 MB | NPU |
| YamNet | QNN_DLC | float | Qualcomm® SA8255P | 0.36 ms | 0 - 22 MB | NPU |
| YamNet | QNN_DLC | float | Qualcomm® QCS8450 | 0.35 ms | 0 - 43 MB | NPU |
| YamNet | QNN_DLC | float | Qualcomm® Dragonwing™ IQ-9075 | 0.273 ms | 0 - 2 MB | NPU |
| YamNet | QNN_DLC | float | Qualcomm® Dragonwing™ IQ-X7181 | 0.278 ms | 0 - 0 MB | NPU |
| YamNet | QNN_DLC | float | Qualcomm® Dragonwing™ Q-8750 | 0.154 ms | 0 - 19 MB | NPU |
| YamNet | QNN_DLC | float | Qualcomm® SA7255P | 0.678 ms | 0 - 21 MB | NPU |
| YamNet | QNN_DLC | float | Qualcomm® SA8295P | 0.561 ms | 0 - 19 MB | NPU |
| YamNet | QNN_DLC | float | Snapdragon® 8 Elite Mobile | 0.154 ms | 0 - 19 MB | NPU |
| YamNet | QNN_DLC | float | Snapdragon® 8 Elite Gen 5 Mobile | 0.154 ms | 0 - 20 MB | NPU |
| YamNet | QNN_DLC | w8a16 | Snapdragon® X2 Elite | 0.177 ms | 1 - 1 MB | NPU |
| YamNet | QNN_DLC | w8a16 | Snapdragon® X Elite | 0.195 ms | 0 - 0 MB | NPU |
| YamNet | QNN_DLC | w8a16 | Snapdragon® 8 Gen 3 Mobile | 0.1 ms | 0 - 37 MB | NPU |
| YamNet | QNN_DLC | w8a16 | Snapdragon® 8 Gen 1 Mobile | 0.215 ms | 0 - 42 MB | NPU |
| YamNet | QNN_DLC | w8a16 | Qualcomm® Dragonwing™ QCS6490 | 0.663 ms | 0 - 2 MB | NPU |
| YamNet | QNN_DLC | w8a16 | Qualcomm® Dragonwing™ IQ-8275 | 0.188 ms | 0 - 2 MB | NPU |
| YamNet | QNN_DLC | w8a16 | Qualcomm® Dragonwing™ IQ-8275 | 0.421 ms | 0 - 25 MB | NPU |
| YamNet | QNN_DLC | w8a16 | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 0.133 ms | 0 - 2 MB | NPU |
| YamNet | QNN_DLC | w8a16 | Qualcomm® SA8775P | 0.28 ms | 0 - 26 MB | NPU |
| YamNet | QNN_DLC | w8a16 | Qualcomm® SA8650P | 0.28 ms | 0 - 26 MB | NPU |
| YamNet | QNN_DLC | w8a16 | Qualcomm® SA8255P | 0.28 ms | 0 - 26 MB | NPU |
| YamNet | QNN_DLC | w8a16 | Qualcomm® QCS8450 | 0.215 ms | 0 - 42 MB | NPU |
| YamNet | QNN_DLC | w8a16 | Qualcomm® Dragonwing™ IQ-9075 | 0.189 ms | 0 - 2 MB | NPU |
| YamNet | QNN_DLC | w8a16 | Qualcomm® Dragonwing™ IQ-X7181 | 0.195 ms | 0 - 0 MB | NPU |
| YamNet | QNN_DLC | w8a16 | Qualcomm® Dragonwing™ Q-6690 | 0.504 ms | 0 - 24 MB | NPU |
| YamNet | QNN_DLC | w8a16 | Qualcomm® Dragonwing™ Q-7790 | 0.165 ms | 0 - 24 MB | NPU |
| YamNet | QNN_DLC | w8a16 | Qualcomm® Dragonwing™ Q-8750 | 0.097 ms | 0 - 23 MB | NPU |
| YamNet | QNN_DLC | w8a16 | Qualcomm® SA7255P | 0.421 ms | 0 - 25 MB | NPU |
| YamNet | QNN_DLC | w8a16 | Qualcomm® SA8295P | 0.435 ms | 0 - 22 MB | NPU |
| YamNet | QNN_DLC | w8a16 | Snapdragon® 8 Elite Mobile | 0.097 ms | 0 - 23 MB | NPU |
| YamNet | QNN_DLC | w8a16 | Snapdragon® 8 Elite Gen 5 Mobile | 0.092 ms | 0 - 24 MB | NPU |
| YamNet | QNN_DLC | w8a16 | Snapdragon® 7 Gen 4 Mobile | 0.165 ms | 0 - 24 MB | NPU |
| YamNet | QNN_DLC | w8a8 | Snapdragon® X2 Elite | 0.156 ms | 1 - 1 MB | NPU |
| YamNet | QNN_DLC | w8a8 | Snapdragon® X Elite | 0.21 ms | 1 - 1 MB | NPU |
| YamNet | QNN_DLC | w8a8 | Snapdragon® 8 Gen 3 Mobile | 0.113 ms | 0 - 36 MB | NPU |
| YamNet | QNN_DLC | w8a8 | Snapdragon® 8 Gen 1 Mobile | 0.206 ms | 0 - 42 MB | NPU |
| YamNet | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ QCS6490 | 0.558 ms | 2 - 4 MB | NPU |
| YamNet | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ IQ-8275 | 0.182 ms | 0 - 2 MB | NPU |
| YamNet | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ IQ-8275 | 0.386 ms | 0 - 27 MB | NPU |
| YamNet | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 0.13 ms | 0 - 1 MB | NPU |
| YamNet | QNN_DLC | w8a8 | Qualcomm® SA8775P | 0.272 ms | 0 - 26 MB | NPU |
| YamNet | QNN_DLC | w8a8 | Qualcomm® SA8650P | 0.272 ms | 0 - 26 MB | NPU |
| YamNet | QNN_DLC | w8a8 | Qualcomm® SA8255P | 0.272 ms | 0 - 26 MB | NPU |
| YamNet | QNN_DLC | w8a8 | Qualcomm® QCS8450 | 0.206 ms | 0 - 42 MB | NPU |
| YamNet | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ IQ-9075 | 0.187 ms | 2 - 4 MB | NPU |
| YamNet | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ IQ-X7181 | 0.21 ms | 1 - 1 MB | NPU |
| YamNet | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ Q-6690 | 0.434 ms | 0 - 24 MB | NPU |
| YamNet | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ Q-7790 | 0.159 ms | 0 - 24 MB | NPU |
| YamNet | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ Q-8750 | 0.1 ms | 0 - 23 MB | NPU |
| YamNet | QNN_DLC | w8a8 | Qualcomm® SA7255P | 0.386 ms | 0 - 27 MB | NPU |
| YamNet | QNN_DLC | w8a8 | Qualcomm® SA8295P | 0.434 ms | 0 - 22 MB | NPU |
| YamNet | QNN_DLC | w8a8 | Snapdragon® 8 Elite Mobile | 0.1 ms | 0 - 23 MB | NPU |
| YamNet | QNN_DLC | w8a8 | Snapdragon® 8 Elite Gen 5 Mobile | 0.094 ms | 0 - 23 MB | NPU |
| YamNet | QNN_DLC | w8a8 | Snapdragon® 7 Gen 4 Mobile | 0.159 ms | 0 - 24 MB | NPU |
| YamNet | TFLITE | float | Snapdragon® 8 Gen 3 Mobile | 0.181 ms | 0 - 42 MB | NPU |
| YamNet | TFLITE | float | Snapdragon® 8 Gen 1 Mobile | 0.351 ms | 0 - 48 MB | NPU |
| YamNet | TFLITE | float | Qualcomm® Dragonwing™ IQ-8275 | 0.297 ms | 0 - 10 MB | NPU |
| YamNet | TFLITE | float | Qualcomm® Dragonwing™ IQ-8275 | 0.69 ms | 0 - 28 MB | NPU |
| YamNet | TFLITE | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 0.211 ms | 0 - 7 MB | NPU |
| YamNet | TFLITE | float | Qualcomm® SA8775P | 0.374 ms | 0 - 28 MB | NPU |
| YamNet | TFLITE | float | Qualcomm® SA8650P | 0.374 ms | 0 - 28 MB | NPU |
| YamNet | TFLITE | float | Qualcomm® SA8255P | 0.374 ms | 0 - 28 MB | NPU |
| YamNet | TFLITE | float | Qualcomm® QCS8450 | 0.351 ms | 0 - 48 MB | NPU |
| YamNet | TFLITE | float | Qualcomm® Dragonwing™ IQ-9075 | 0.277 ms | 0 - 10 MB | NPU |
| YamNet | TFLITE | float | Qualcomm® Dragonwing™ Q-8750 | 0.155 ms | 0 - 26 MB | NPU |
| YamNet | TFLITE | float | Qualcomm® SA7255P | 0.69 ms | 0 - 28 MB | NPU |
| YamNet | TFLITE | float | Qualcomm® SA8295P | 0.578 ms | 0 - 25 MB | NPU |
| YamNet | TFLITE | float | Snapdragon® 8 Elite Mobile | 0.155 ms | 0 - 26 MB | NPU |
| YamNet | TFLITE | float | Snapdragon® 8 Elite Gen 5 Mobile | 0.155 ms | 0 - 28 MB | NPU |
| YamNet | TFLITE | w8a8 | Snapdragon® 8 Gen 3 Mobile | 0.109 ms | 0 - 36 MB | NPU |
| YamNet | TFLITE | w8a8 | Snapdragon® 8 Gen 1 Mobile | 0.205 ms | 0 - 42 MB | NPU |
| YamNet | TFLITE | w8a8 | Qualcomm® Dragonwing™ QCS6490 | 0.525 ms | 0 - 6 MB | NPU |
| YamNet | TFLITE | w8a8 | Qualcomm® Dragonwing™ IQ-8275 | 0.19 ms | 0 - 6 MB | NPU |
| YamNet | TFLITE | w8a8 | Qualcomm® Dragonwing™ IQ-8275 | 0.414 ms | 0 - 24 MB | NPU |
| YamNet | TFLITE | w8a8 | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 0.132 ms | 0 - 7 MB | NPU |
| YamNet | TFLITE | w8a8 | Qualcomm® SA8775P | 0.287 ms | 0 - 25 MB | NPU |
| YamNet | TFLITE | w8a8 | Qualcomm® SA8650P | 0.287 ms | 0 - 25 MB | NPU |
| YamNet | TFLITE | w8a8 | Qualcomm® SA8255P | 0.287 ms | 0 - 25 MB | NPU |
| YamNet | TFLITE | w8a8 | Qualcomm® QCS8450 | 0.205 ms | 0 - 42 MB | NPU |
| YamNet | TFLITE | w8a8 | Qualcomm® Dragonwing™ IQ-9075 | 0.189 ms | 0 - 6 MB | NPU |
| YamNet | TFLITE | w8a8 | Qualcomm® Dragonwing™ Q-6690 | 0.432 ms | 0 - 22 MB | NPU |
| YamNet | TFLITE | w8a8 | Qualcomm® Dragonwing™ Q-7790 | 0.17 ms | 0 - 22 MB | NPU |
| YamNet | TFLITE | w8a8 | Qualcomm® Dragonwing™ Q-8750 | 0.1 ms | 0 - 27 MB | NPU |
| YamNet | TFLITE | w8a8 | Qualcomm® SA7255P | 0.414 ms | 0 - 24 MB | NPU |
| YamNet | TFLITE | w8a8 | Qualcomm® SA8295P | 0.437 ms | 0 - 21 MB | NPU |
| YamNet | TFLITE | w8a8 | Snapdragon® 8 Elite Mobile | 0.1 ms | 0 - 27 MB | NPU |
| YamNet | TFLITE | w8a8 | Snapdragon® 8 Elite Gen 5 Mobile | 0.098 ms | 0 - 23 MB | NPU |
| YamNet | TFLITE | w8a8 | Snapdragon® 7 Gen 4 Mobile | 0.17 ms | 0 - 22 MB | NPU |
License
- The license for the original implementation of YamNet can be found here.
References
- MobileNets Efficient Convolutional Neural Networks for Mobile Vision Applications
- Source Model Implementation
Community
- Join our AI Hub Slack community to collaborate, post questions and learn more about on-device AI.
- For questions or feedback please reach out to us.
