ResNet18: Optimized for Qualcomm Devices
ResNet18 is a machine learning model that can classify images from the Imagenet dataset. It can also be used as a backbone in building more complex models for specific use cases.
This is based on the implementation of ResNet18 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 | w8a8 | Universal | QAIRT 2.45, ONNX Runtime 1.27.1 | Download |
| QNN_DLC | float | 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 ResNet18 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 ResNet18 on GitHub for usage instructions.
Model Details
Model Type: Model_use_case.image_classification
Model Stats:
- Input resolution: 224x224
- Model checkpoint: Imagenet
- Model size (float): 44.6 MB
- Model size (w8a8): 11.3 MB
- Number of parameters: 11.7M
Performance Summary
| Model | Runtime | Precision | Chipset | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit |
|---|---|---|---|---|---|---|
| ResNet18 | ONNX | float | Snapdragon® X2 Elite | 0.509 ms | 2 - 2 MB | NPU |
| ResNet18 | ONNX | float | Snapdragon® X Elite | 1.022 ms | 23 - 23 MB | NPU |
| ResNet18 | ONNX | float | Snapdragon® 8 Gen 3 Mobile | 0.782 ms | 0 - 39 MB | NPU |
| ResNet18 | ONNX | float | Snapdragon® 8 Gen 1 Mobile | 1.837 ms | 1 - 40 MB | NPU |
| ResNet18 | ONNX | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 1.049 ms | 0 - 34 MB | NPU |
| ResNet18 | ONNX | float | Qualcomm® QCS8450 | 1.837 ms | 1 - 40 MB | NPU |
| ResNet18 | ONNX | float | Qualcomm® Dragonwing™ IQ-9075 | 1.693 ms | 1 - 3 MB | NPU |
| ResNet18 | ONNX | float | Qualcomm® Dragonwing™ IQ-X7181 | 1.022 ms | 23 - 23 MB | NPU |
| ResNet18 | ONNX | float | Qualcomm® Dragonwing™ Q-8750 | 0.616 ms | 0 - 22 MB | NPU |
| ResNet18 | ONNX | float | Snapdragon® 8 Elite Mobile | 0.616 ms | 0 - 22 MB | NPU |
| ResNet18 | ONNX | float | Snapdragon® 8 Elite Gen 5 Mobile | 0.556 ms | 0 - 27 MB | NPU |
| ResNet18 | ONNX | w8a8 | Snapdragon® X2 Elite | 0.262 ms | 1 - 1 MB | NPU |
| ResNet18 | ONNX | w8a8 | Snapdragon® X Elite | 0.495 ms | 12 - 12 MB | NPU |
| ResNet18 | ONNX | w8a8 | Snapdragon® 8 Gen 3 Mobile | 0.414 ms | 0 - 50 MB | NPU |
| ResNet18 | ONNX | w8a8 | Snapdragon® 8 Gen 1 Mobile | 0.75 ms | 0 - 55 MB | NPU |
| ResNet18 | ONNX | w8a8 | Qualcomm® Dragonwing™ QCS6490 | 1.868 ms | 0 - 3 MB | NPU |
| ResNet18 | ONNX | w8a8 | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 0.554 ms | 0 - 124 MB | NPU |
| ResNet18 | ONNX | w8a8 | Qualcomm® QCS8450 | 0.75 ms | 0 - 55 MB | NPU |
| ResNet18 | ONNX | w8a8 | Qualcomm® Dragonwing™ IQ-9075 | 0.661 ms | 0 - 3 MB | NPU |
| ResNet18 | ONNX | w8a8 | Qualcomm® Dragonwing™ IQ-X7181 | 0.495 ms | 12 - 12 MB | NPU |
| ResNet18 | ONNX | w8a8 | Qualcomm® Dragonwing™ Q-8750 | 0.361 ms | 0 - 24 MB | NPU |
| ResNet18 | ONNX | w8a8 | Snapdragon® 8 Elite Mobile | 0.361 ms | 0 - 24 MB | NPU |
| ResNet18 | ONNX | w8a8 | Snapdragon® 8 Elite Gen 5 Mobile | 0.344 ms | 0 - 25 MB | NPU |
| ResNet18 | QNN_DLC | float | Snapdragon® X2 Elite | 0.663 ms | 1 - 1 MB | NPU |
| ResNet18 | QNN_DLC | float | Snapdragon® X Elite | 1.449 ms | 1 - 1 MB | NPU |
| ResNet18 | QNN_DLC | float | Snapdragon® 8 Gen 3 Mobile | 0.92 ms | 0 - 39 MB | NPU |
| ResNet18 | QNN_DLC | float | Snapdragon® 8 Gen 1 Mobile | 2.43 ms | 0 - 37 MB | NPU |
| ResNet18 | QNN_DLC | float | Qualcomm® Dragonwing™ QCS8275 | 5.875 ms | 1 - 24 MB | NPU |
| ResNet18 | QNN_DLC | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 1.323 ms | 1 - 14 MB | NPU |
| ResNet18 | QNN_DLC | float | Qualcomm® SA8775P | 1.944 ms | 1 - 26 MB | NPU |
| ResNet18 | QNN_DLC | float | Qualcomm® SA8650P | 1.944 ms | 1 - 26 MB | NPU |
| ResNet18 | QNN_DLC | float | Qualcomm® SA8255P | 1.944 ms | 1 - 26 MB | NPU |
| ResNet18 | QNN_DLC | float | Qualcomm® QCS8450 | 2.43 ms | 0 - 37 MB | NPU |
| ResNet18 | QNN_DLC | float | Qualcomm® Dragonwing™ IQ-9075 | 1.813 ms | 1 - 3 MB | NPU |
| ResNet18 | QNN_DLC | float | Qualcomm® Dragonwing™ IQ-X7181 | 1.449 ms | 1 - 1 MB | NPU |
| ResNet18 | QNN_DLC | float | Qualcomm® Dragonwing™ Q-8750 | 0.715 ms | 1 - 28 MB | NPU |
| ResNet18 | QNN_DLC | float | Qualcomm® SA7255P | 5.875 ms | 1 - 24 MB | NPU |
| ResNet18 | QNN_DLC | float | Qualcomm® SA8295P | 2.315 ms | 0 - 20 MB | NPU |
| ResNet18 | QNN_DLC | float | Snapdragon® 8 Elite Mobile | 0.715 ms | 1 - 28 MB | NPU |
| ResNet18 | QNN_DLC | float | Snapdragon® 8 Elite Gen 5 Mobile | 0.593 ms | 1 - 24 MB | NPU |
| ResNet18 | QNN_DLC | w8a8 | Snapdragon® X2 Elite | 0.313 ms | 0 - 0 MB | NPU |
| ResNet18 | QNN_DLC | w8a8 | Snapdragon® X Elite | 0.617 ms | 0 - 0 MB | NPU |
| ResNet18 | QNN_DLC | w8a8 | Snapdragon® 8 Gen 3 Mobile | 0.412 ms | 0 - 47 MB | NPU |
| ResNet18 | QNN_DLC | w8a8 | Snapdragon® 8 Gen 1 Mobile | 0.744 ms | 0 - 54 MB | NPU |
| ResNet18 | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ QCS6490 | 2.051 ms | 2 - 4 MB | NPU |
| ResNet18 | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ QCS8275 | 1.317 ms | 0 - 24 MB | NPU |
| ResNet18 | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 0.543 ms | 0 - 1 MB | NPU |
| ResNet18 | QNN_DLC | w8a8 | Qualcomm® SA8775P | 0.718 ms | 0 - 26 MB | NPU |
| ResNet18 | QNN_DLC | w8a8 | Qualcomm® SA8650P | 0.718 ms | 0 - 26 MB | NPU |
| ResNet18 | QNN_DLC | w8a8 | Qualcomm® SA8255P | 0.718 ms | 0 - 26 MB | NPU |
| ResNet18 | QNN_DLC | w8a8 | Qualcomm® QCS8450 | 0.744 ms | 0 - 54 MB | NPU |
| ResNet18 | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ IQ-9075 | 0.616 ms | 2 - 4 MB | NPU |
| ResNet18 | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ IQ-X7181 | 0.617 ms | 0 - 0 MB | NPU |
| ResNet18 | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ Q-6690 | 3.232 ms | 0 - 33 MB | NPU |
| ResNet18 | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ Q-7790 | 0.681 ms | 0 - 32 MB | NPU |
| ResNet18 | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ Q-8750 | 0.313 ms | 0 - 27 MB | NPU |
| ResNet18 | QNN_DLC | w8a8 | Qualcomm® SA7255P | 1.317 ms | 0 - 24 MB | NPU |
| ResNet18 | QNN_DLC | w8a8 | Qualcomm® SA8295P | 0.947 ms | 0 - 22 MB | NPU |
| ResNet18 | QNN_DLC | w8a8 | Snapdragon® 8 Elite Mobile | 0.313 ms | 0 - 27 MB | NPU |
| ResNet18 | QNN_DLC | w8a8 | Snapdragon® 8 Elite Gen 5 Mobile | 0.276 ms | 0 - 24 MB | NPU |
| ResNet18 | QNN_DLC | w8a8 | Snapdragon® 7 Gen 4 Mobile | 0.681 ms | 0 - 32 MB | NPU |
| ResNet18 | TFLITE | float | Snapdragon® 8 Gen 3 Mobile | 0.913 ms | 0 - 55 MB | NPU |
| ResNet18 | TFLITE | float | Snapdragon® 8 Gen 1 Mobile | 2.465 ms | 0 - 58 MB | NPU |
| ResNet18 | TFLITE | float | Qualcomm® Dragonwing™ QCS8275 | 5.845 ms | 0 - 27 MB | NPU |
| ResNet18 | TFLITE | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 1.319 ms | 0 - 16 MB | NPU |
| ResNet18 | TFLITE | float | Qualcomm® SA8775P | 1.916 ms | 0 - 28 MB | NPU |
| ResNet18 | TFLITE | float | Qualcomm® SA8650P | 1.916 ms | 0 - 28 MB | NPU |
| ResNet18 | TFLITE | float | Qualcomm® SA8255P | 1.916 ms | 0 - 28 MB | NPU |
| ResNet18 | TFLITE | float | Qualcomm® QCS8450 | 2.465 ms | 0 - 58 MB | NPU |
| ResNet18 | TFLITE | float | Qualcomm® Dragonwing™ IQ-9075 | 2.013 ms | 0 - 25 MB | NPU |
| ResNet18 | TFLITE | float | Qualcomm® Dragonwing™ Q-8750 | 0.711 ms | 0 - 28 MB | NPU |
| ResNet18 | TFLITE | float | Qualcomm® SA7255P | 5.845 ms | 0 - 27 MB | NPU |
| ResNet18 | TFLITE | float | Qualcomm® SA8295P | 2.262 ms | 0 - 22 MB | NPU |
| ResNet18 | TFLITE | float | Snapdragon® 8 Elite Mobile | 0.711 ms | 0 - 28 MB | NPU |
| ResNet18 | TFLITE | float | Snapdragon® 8 Elite Gen 5 Mobile | 0.589 ms | 0 - 26 MB | NPU |
| ResNet18 | TFLITE | w8a8 | Snapdragon® 8 Gen 3 Mobile | 0.292 ms | 0 - 47 MB | NPU |
| ResNet18 | TFLITE | w8a8 | Snapdragon® 8 Gen 1 Mobile | 0.607 ms | 0 - 53 MB | NPU |
| ResNet18 | TFLITE | w8a8 | Qualcomm® Dragonwing™ QCS6490 | 1.639 ms | 0 - 13 MB | NPU |
| ResNet18 | TFLITE | w8a8 | Qualcomm® Dragonwing™ QCS8275 | 1.01 ms | 0 - 24 MB | NPU |
| ResNet18 | TFLITE | w8a8 | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 0.383 ms | 0 - 1 MB | NPU |
| ResNet18 | TFLITE | w8a8 | Qualcomm® SA8775P | 0.563 ms | 0 - 25 MB | NPU |
| ResNet18 | TFLITE | w8a8 | Qualcomm® SA8650P | 0.563 ms | 0 - 25 MB | NPU |
| ResNet18 | TFLITE | w8a8 | Qualcomm® SA8255P | 0.563 ms | 0 - 25 MB | NPU |
| ResNet18 | TFLITE | w8a8 | Qualcomm® QCS8450 | 0.607 ms | 0 - 53 MB | NPU |
| ResNet18 | TFLITE | w8a8 | Qualcomm® Dragonwing™ IQ-9075 | 0.469 ms | 0 - 13 MB | NPU |
| ResNet18 | TFLITE | w8a8 | Qualcomm® Dragonwing™ Q-6690 | 2.796 ms | 0 - 31 MB | NPU |
| ResNet18 | TFLITE | w8a8 | Qualcomm® Dragonwing™ Q-7790 | 0.545 ms | 0 - 31 MB | NPU |
| ResNet18 | TFLITE | w8a8 | Qualcomm® Dragonwing™ Q-8750 | 0.25 ms | 0 - 23 MB | NPU |
| ResNet18 | TFLITE | w8a8 | Qualcomm® SA7255P | 1.01 ms | 0 - 24 MB | NPU |
| ResNet18 | TFLITE | w8a8 | Qualcomm® SA8295P | 0.766 ms | 0 - 22 MB | NPU |
| ResNet18 | TFLITE | w8a8 | Snapdragon® 8 Elite Mobile | 0.25 ms | 0 - 23 MB | NPU |
| ResNet18 | TFLITE | w8a8 | Snapdragon® 8 Elite Gen 5 Mobile | 0.23 ms | 0 - 24 MB | NPU |
| ResNet18 | TFLITE | w8a8 | Snapdragon® 7 Gen 4 Mobile | 0.545 ms | 0 - 31 MB | NPU |
License
- The license for the original implementation of ResNet18 can be found here.
References
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.
