--- library_name: pytorch license: other tags: - backbone - bu_auto - android pipeline_tag: image-classification --- ![](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/efficientnet_b4/web-assets/model_demo.png) # EfficientNet-B4: Optimized for Qualcomm Devices EfficientNetB4 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-B4 found [here](https://github.com/pytorch/vision/blob/main/torchvision/models/efficientnet.py). 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/main/src/qai_hub_models/models/efficientnet_b4) 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 2.42, ONNX Runtime 1.25.0 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/efficientnet_b4/releases/v0.55.0/efficientnet_b4-onnx-float.zip) | QNN_DLC | float | Universal | QAIRT 2.45 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/efficientnet_b4/releases/v0.55.0/efficientnet_b4-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/efficientnet_b4/releases/v0.55.0/efficientnet_b4-qnn_dlc-w8a16.zip) | TFLITE | float | Universal | QAIRT 2.45 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/efficientnet_b4/releases/v0.55.0/efficientnet_b4-tflite-float.zip) For more device-specific assets and performance metrics, visit **[EfficientNet-B4 on Qualcomm® AI Hub](https://aihub.qualcomm.com/models/efficientnet_b4)**. ### Option 2: Export with Custom Configurations Use the [Qualcomm® AI Hub Models](https://github.com/qualcomm/ai-hub-models/blob/main/src/qai_hub_models/models/efficientnet_b4) 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-B4 on GitHub](https://github.com/qualcomm/ai-hub-models/blob/main/src/qai_hub_models/models/efficientnet_b4) for usage instructions. ## Model Details **Model Type:** Model_use_case.image_classification **Model Stats:** - Model checkpoint: Imagenet - Input resolution: 380x380 - Number of parameters: 19.3M - Model size (float): 73.6 MB - Model size (w8a16): 24.0 MB ## Performance Summary | Model | Runtime | Precision | Chipset | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit |---|---|---|---|---|---|--- | EfficientNet-B4 | ONNX | float | Snapdragon® 8 Elite Gen 5 Mobile | 3.01 ms | 1 - 209 MB | NPU | EfficientNet-B4 | ONNX | float | Snapdragon® X2 Elite | 3.926 ms | 210 - 210 MB | NPU | EfficientNet-B4 | ONNX | float | Snapdragon® X Elite | 7.733 ms | 147 - 147 MB | NPU | EfficientNet-B4 | ONNX | float | Snapdragon® 8 Gen 3 Mobile | 5.346 ms | 25 - 180 MB | NPU | EfficientNet-B4 | ONNX | float | Qualcomm® QCS8550 (Proxy) | 7.374 ms | 2 - 95 MB | NPU | EfficientNet-B4 | ONNX | float | Snapdragon® 8 Elite For Galaxy Mobile | 4.065 ms | 0 - 93 MB | NPU | EfficientNet-B4 | ONNX | float | Qualcomm® QCS9075 | 10.766 ms | 2 - 47 MB | NPU | EfficientNet-B4 | ONNX | float | Qualcomm® QCS8750 | 4.065 ms | 0 - 93 MB | NPU | EfficientNet-B4 | ONNX | float | Qualcomm® QCS7181 | 7.733 ms | 147 - 147 MB | NPU | EfficientNet-B4 | QNN_DLC | float | Snapdragon® 8 Elite Gen 5 Mobile | 3.243 ms | 2 - 205 MB | NPU | EfficientNet-B4 | QNN_DLC | float | Snapdragon® X2 Elite | 4.507 ms | 2 - 2 MB | NPU | EfficientNet-B4 | QNN_DLC | float | Snapdragon® X Elite | 8.904 ms | 2 - 2 MB | NPU | EfficientNet-B4 | QNN_DLC | float | Snapdragon® 8 Gen 3 Mobile | 5.813 ms | 0 - 142 MB | NPU | EfficientNet-B4 | QNN_DLC | float | Qualcomm® QCS8275 | 29.139 ms | 2 - 81 MB | NPU | EfficientNet-B4 | QNN_DLC | float | Qualcomm® QCS8550 (Proxy) | 8.122 ms | 2 - 4 MB | NPU | EfficientNet-B4 | QNN_DLC | float | Qualcomm® SA8775P | 10.306 ms | 2 - 85 MB | NPU | EfficientNet-B4 | QNN_DLC | float | Qualcomm® SA8650P | 10.306 ms | 2 - 85 MB | NPU | EfficientNet-B4 | QNN_DLC | float | Qualcomm® SA8255P | 10.306 ms | 2 - 85 MB | NPU | EfficientNet-B4 | QNN_DLC | float | Qualcomm® QCS8450 (Proxy) | 23.062 ms | 0 - 185 MB | NPU | EfficientNet-B4 | QNN_DLC | float | Qualcomm® SA7255P | 29.139 ms | 2 - 81 MB | NPU | EfficientNet-B4 | QNN_DLC | float | Qualcomm® SA8295P | 18.807 ms | 2 - 125 MB | NPU | EfficientNet-B4 | QNN_DLC | float | Snapdragon® 8 Elite For Galaxy Mobile | 4.308 ms | 0 - 85 MB | NPU | EfficientNet-B4 | QNN_DLC | float | Qualcomm® QCS9075 | 12.186 ms | 2 - 5 MB | NPU | EfficientNet-B4 | QNN_DLC | float | Qualcomm® QCS8750 | 4.308 ms | 0 - 85 MB | NPU | EfficientNet-B4 | QNN_DLC | float | Qualcomm® QCS7181 | 8.904 ms | 2 - 2 MB | NPU | EfficientNet-B4 | QNN_DLC | w8a16 | Snapdragon® 8 Elite Gen 5 Mobile | 3.024 ms | 1 - 155 MB | NPU | EfficientNet-B4 | QNN_DLC | w8a16 | Snapdragon® X2 Elite | 3.547 ms | 1 - 1 MB | NPU | EfficientNet-B4 | QNN_DLC | w8a16 | Snapdragon® X Elite | 9.056 ms | 1 - 1 MB | NPU | EfficientNet-B4 | QNN_DLC | w8a16 | Snapdragon® 8 Gen 3 Mobile | 5.67 ms | 1 - 199 MB | NPU | EfficientNet-B4 | QNN_DLC | w8a16 | Qualcomm® QCS6490 | 22.892 ms | 3 - 5 MB | NPU | EfficientNet-B4 | QNN_DLC | w8a16 | Qualcomm® QCS8275 | 15.515 ms | 1 - 141 MB | NPU | EfficientNet-B4 | QNN_DLC | w8a16 | Qualcomm® QCS8550 (Proxy) | 8.38 ms | 1 - 3 MB | NPU | EfficientNet-B4 | QNN_DLC | w8a16 | Qualcomm® SA8775P | 8.934 ms | 1 - 144 MB | NPU | EfficientNet-B4 | QNN_DLC | w8a16 | Qualcomm® SA8650P | 8.934 ms | 1 - 144 MB | NPU | EfficientNet-B4 | QNN_DLC | w8a16 | Qualcomm® SA8255P | 8.934 ms | 1 - 144 MB | NPU | EfficientNet-B4 | QNN_DLC | w8a16 | Qualcomm® QCM6690 | 47.264 ms | 1 - 275 MB | NPU | EfficientNet-B4 | QNN_DLC | w8a16 | Qualcomm® SA7255P | 15.515 ms | 1 - 141 MB | NPU | EfficientNet-B4 | QNN_DLC | w8a16 | Qualcomm® SA8295P | 10.955 ms | 1 - 143 MB | NPU | EfficientNet-B4 | QNN_DLC | w8a16 | Snapdragon® 7 Gen 4 Mobile | 9.702 ms | 1 - 269 MB | NPU | EfficientNet-B4 | QNN_DLC | w8a16 | Snapdragon® 8 Elite For Galaxy Mobile | 3.753 ms | 1 - 146 MB | NPU | EfficientNet-B4 | QNN_DLC | w8a16 | Qualcomm® QCS9075 | 9.888 ms | 0 - 3 MB | NPU | EfficientNet-B4 | QNN_DLC | w8a16 | Qualcomm® QCS8450 (Proxy) | 11.26 ms | 0 - 201 MB | NPU | EfficientNet-B4 | QNN_DLC | w8a16 | Qualcomm® QCS7790 | 9.702 ms | 1 - 269 MB | NPU | EfficientNet-B4 | QNN_DLC | w8a16 | Qualcomm® QCS8750 | 3.753 ms | 1 - 146 MB | NPU | EfficientNet-B4 | QNN_DLC | w8a16 | Qualcomm® QCS7181 | 9.056 ms | 1 - 1 MB | NPU | EfficientNet-B4 | TFLITE | float | Snapdragon® 8 Elite Gen 5 Mobile | 3.202 ms | 0 - 102 MB | NPU | EfficientNet-B4 | TFLITE | float | Snapdragon® 8 Gen 3 Mobile | 5.749 ms | 0 - 160 MB | NPU | EfficientNet-B4 | TFLITE | float | Qualcomm® QCS8275 | 28.908 ms | 0 - 97 MB | NPU | EfficientNet-B4 | TFLITE | float | Qualcomm® QCS8550 (Proxy) | 8.032 ms | 0 - 2 MB | NPU | EfficientNet-B4 | TFLITE | float | Qualcomm® SA8775P | 10.264 ms | 0 - 99 MB | NPU | EfficientNet-B4 | TFLITE | float | Qualcomm® SA8650P | 10.264 ms | 0 - 99 MB | NPU | EfficientNet-B4 | TFLITE | float | Qualcomm® SA8255P | 10.264 ms | 0 - 99 MB | NPU | EfficientNet-B4 | TFLITE | float | Qualcomm® QCS8450 (Proxy) | 21.932 ms | 0 - 202 MB | NPU | EfficientNet-B4 | TFLITE | float | Qualcomm® SA7255P | 28.908 ms | 0 - 97 MB | NPU | EfficientNet-B4 | TFLITE | float | Qualcomm® SA8295P | 18.861 ms | 0 - 140 MB | NPU | EfficientNet-B4 | TFLITE | float | Snapdragon® 8 Elite For Galaxy Mobile | 4.319 ms | 0 - 105 MB | NPU | EfficientNet-B4 | TFLITE | float | Qualcomm® QCS9075 | 10.99 ms | 0 - 49 MB | NPU | EfficientNet-B4 | TFLITE | float | Qualcomm® QCS8750 | 4.319 ms | 0 - 105 MB | NPU ## License * The license for the original implementation of EfficientNet-B4 can be found [here](https://github.com/pytorch/vision/blob/main/LICENSE). ## References * [EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks](https://arxiv.org/abs/1905.11946) * [Source Model Implementation](https://github.com/pytorch/vision/blob/main/torchvision/models/efficientnet.py) ## 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).