EfficientNetV2: Smaller Models and Faster Training
Paper • 2104.00298 • Published • 1
EfficientNet-V2-s: Optimized for Qualcomm Devices
EfficientNetV2-s 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 EfficientNet-V2-s 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 2.42, ONNX Runtime 1.24.3 | Download |
| ONNX | w8a16 | Universal | QAIRT 2.42, ONNX Runtime 1.24.3 | Download |
| QNN_DLC | float | Universal | QAIRT 2.45 | Download |
| QNN_DLC | w8a16 | Universal | QAIRT 2.45 | Download |
| TFLITE | float | Universal | QAIRT 2.45 | Download |
For more device-specific assets and performance metrics, visit EfficientNet-V2-s 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 EfficientNet-V2-s on GitHub for usage instructions.
Model Details
Model Type: Model_use_case.image_classification
Model Stats:
- Model checkpoint: Imagenet
- Input resolution: 384x384
- Number of parameters: 21.4M
- Model size (float): 81.7 MB
- Model size (w8a16): 27.2 MB
Performance Summary
| Model | Runtime | Precision | Chipset | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit |
|---|---|---|---|---|---|---|
| EfficientNet-V2-s | ONNX | float | Snapdragon® 8 Elite Gen 5 Mobile | 1.186 ms | 0 - 76 MB | NPU |
| EfficientNet-V2-s | ONNX | float | Snapdragon® 8 Elite Mobile | 1.425 ms | 0 - 70 MB | NPU |
| EfficientNet-V2-s | ONNX | float | Snapdragon® X2 Elite | 1.318 ms | 47 - 47 MB | NPU |
| EfficientNet-V2-s | ONNX | float | Snapdragon® X Elite | 2.694 ms | 46 - 46 MB | NPU |
| EfficientNet-V2-s | ONNX | float | Snapdragon® X Elite | 2.694 ms | 46 - 46 MB | NPU |
| EfficientNet-V2-s | ONNX | float | Snapdragon® 8 Gen 3 Mobile | 1.822 ms | 0 - 155 MB | NPU |
| EfficientNet-V2-s | ONNX | float | Qualcomm® QCS8550 (Proxy) | 2.457 ms | 0 - 51 MB | NPU |
| EfficientNet-V2-s | ONNX | float | Qualcomm® QCS9075 | 3.475 ms | 0 - 4 MB | NPU |
| EfficientNet-V2-s | ONNX | float | Snapdragon® 8 Elite For Galaxy Mobile | 1.425 ms | 0 - 70 MB | NPU |
| EfficientNet-V2-s | ONNX | w8a16 | Snapdragon® 8 Elite Gen 5 Mobile | 0.933 ms | 0 - 127 MB | NPU |
| EfficientNet-V2-s | ONNX | w8a16 | Snapdragon® 8 Elite Mobile | 1.143 ms | 0 - 118 MB | NPU |
| EfficientNet-V2-s | ONNX | w8a16 | Snapdragon® X2 Elite | 1.089 ms | 24 - 24 MB | NPU |
| EfficientNet-V2-s | ONNX | w8a16 | Snapdragon® X Elite | 2.664 ms | 24 - 24 MB | NPU |
| EfficientNet-V2-s | ONNX | w8a16 | Snapdragon® X Elite | 2.664 ms | 24 - 24 MB | NPU |
| EfficientNet-V2-s | ONNX | w8a16 | Snapdragon® 8 Gen 3 Mobile | 1.602 ms | 0 - 178 MB | NPU |
| EfficientNet-V2-s | ONNX | w8a16 | Qualcomm® QCS6490 | 279.21 ms | 27 - 32 MB | CPU |
| EfficientNet-V2-s | ONNX | w8a16 | Qualcomm® QCS8550 (Proxy) | 2.357 ms | 0 - 32 MB | NPU |
| EfficientNet-V2-s | ONNX | w8a16 | Qualcomm® QCS9075 | 2.679 ms | 0 - 3 MB | NPU |
| EfficientNet-V2-s | ONNX | w8a16 | Qualcomm® QCM6690 | 124.549 ms | 15 - 28 MB | CPU |
| EfficientNet-V2-s | ONNX | w8a16 | Snapdragon® 8 Elite For Galaxy Mobile | 1.143 ms | 0 - 118 MB | NPU |
| EfficientNet-V2-s | ONNX | w8a16 | Snapdragon® 7 Gen 4 Mobile | 114.644 ms | 25 - 39 MB | CPU |
| EfficientNet-V2-s | ONNX | w8a16 | Snapdragon® 7 Gen 4 Mobile | 114.644 ms | 25 - 39 MB | CPU |
| EfficientNet-V2-s | QNN_DLC | float | Snapdragon® 8 Elite Gen 5 Mobile | 1.194 ms | 1 - 68 MB | NPU |
| EfficientNet-V2-s | QNN_DLC | float | Snapdragon® 8 Elite Mobile | 1.486 ms | 1 - 68 MB | NPU |
| EfficientNet-V2-s | QNN_DLC | float | Snapdragon® X2 Elite | 1.575 ms | 1 - 1 MB | NPU |
| EfficientNet-V2-s | QNN_DLC | float | Snapdragon® X Elite | 2.885 ms | 1 - 1 MB | NPU |
| EfficientNet-V2-s | QNN_DLC | float | Snapdragon® X Elite | 2.885 ms | 1 - 1 MB | NPU |
| EfficientNet-V2-s | QNN_DLC | float | Snapdragon® 8 Gen 3 Mobile | 1.925 ms | 0 - 139 MB | NPU |
| EfficientNet-V2-s | QNN_DLC | float | Qualcomm® QCS8275 (Proxy) | 10.778 ms | 1 - 63 MB | NPU |
| EfficientNet-V2-s | QNN_DLC | float | Qualcomm® QCS8550 (Proxy) | 2.642 ms | 1 - 147 MB | NPU |
| EfficientNet-V2-s | QNN_DLC | float | Qualcomm® QCS9075 | 3.641 ms | 1 - 3 MB | NPU |
| EfficientNet-V2-s | QNN_DLC | float | Qualcomm® QCS8450 (Proxy) | 5.675 ms | 0 - 147 MB | NPU |
| EfficientNet-V2-s | QNN_DLC | float | Snapdragon® 8 Elite For Galaxy Mobile | 1.486 ms | 1 - 68 MB | NPU |
| EfficientNet-V2-s | QNN_DLC | w8a16 | Snapdragon® 8 Elite Gen 5 Mobile | 1.003 ms | 0 - 117 MB | NPU |
| EfficientNet-V2-s | QNN_DLC | w8a16 | Snapdragon® 8 Elite Mobile | 1.237 ms | 0 - 112 MB | NPU |
| EfficientNet-V2-s | QNN_DLC | w8a16 | Snapdragon® X2 Elite | 1.434 ms | 0 - 0 MB | NPU |
| EfficientNet-V2-s | QNN_DLC | w8a16 | Snapdragon® X Elite | 2.923 ms | 0 - 0 MB | NPU |
| EfficientNet-V2-s | QNN_DLC | w8a16 | Snapdragon® X Elite | 2.923 ms | 0 - 0 MB | NPU |
| EfficientNet-V2-s | QNN_DLC | w8a16 | Snapdragon® 8 Gen 3 Mobile | 1.767 ms | 0 - 145 MB | NPU |
| EfficientNet-V2-s | QNN_DLC | w8a16 | Qualcomm® QCS6490 | 6.633 ms | 0 - 2 MB | NPU |
| EfficientNet-V2-s | QNN_DLC | w8a16 | Qualcomm® QCS8275 (Proxy) | 5.356 ms | 0 - 105 MB | NPU |
| EfficientNet-V2-s | QNN_DLC | w8a16 | Qualcomm® QCS8550 (Proxy) | 2.63 ms | 0 - 2 MB | NPU |
| EfficientNet-V2-s | QNN_DLC | w8a16 | Qualcomm® QCS9075 | 2.913 ms | 0 - 2 MB | NPU |
| EfficientNet-V2-s | QNN_DLC | w8a16 | Qualcomm® QCM6690 | 14.096 ms | 0 - 228 MB | NPU |
| EfficientNet-V2-s | QNN_DLC | w8a16 | Qualcomm® QCS8450 (Proxy) | 3.274 ms | 0 - 148 MB | NPU |
| EfficientNet-V2-s | QNN_DLC | w8a16 | Snapdragon® 8 Elite For Galaxy Mobile | 1.237 ms | 0 - 112 MB | NPU |
| EfficientNet-V2-s | QNN_DLC | w8a16 | Snapdragon® 7 Gen 4 Mobile | 2.954 ms | 0 - 110 MB | NPU |
| EfficientNet-V2-s | QNN_DLC | w8a16 | Snapdragon® 7 Gen 4 Mobile | 2.954 ms | 0 - 110 MB | NPU |
| EfficientNet-V2-s | TFLITE | float | Snapdragon® 8 Elite Gen 5 Mobile | 1.194 ms | 0 - 108 MB | NPU |
| EfficientNet-V2-s | TFLITE | float | Snapdragon® 8 Elite Mobile | 1.477 ms | 0 - 109 MB | NPU |
| EfficientNet-V2-s | TFLITE | float | Snapdragon® 8 Gen 3 Mobile | 1.918 ms | 0 - 177 MB | NPU |
| EfficientNet-V2-s | TFLITE | float | Qualcomm® QCS8275 (Proxy) | 10.851 ms | 0 - 104 MB | NPU |
| EfficientNet-V2-s | TFLITE | float | Qualcomm® QCS8550 (Proxy) | 2.632 ms | 0 - 2 MB | NPU |
| EfficientNet-V2-s | TFLITE | float | Qualcomm® QCS9075 | 3.685 ms | 0 - 50 MB | NPU |
| EfficientNet-V2-s | TFLITE | float | Qualcomm® QCS8450 (Proxy) | 5.681 ms | 0 - 191 MB | NPU |
| EfficientNet-V2-s | TFLITE | float | Snapdragon® 8 Elite For Galaxy Mobile | 1.477 ms | 0 - 109 MB | NPU |
License
- The license for the original implementation of EfficientNet-V2-s 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.