README.md

---
library_name: pytorch
license: bsd-3-clause
tags:
- backbone
- bu_auto
- real_time
- android
pipeline_tag: image-classification

---

![](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/efficientvit_b2_cls/web-assets/model_demo.png)

# EfficientViT-b2-cls: Optimized for Qualcomm Devices

EfficientViT 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 EfficientViT-b2-cls found [here](https://github.com/CVHub520/efficientvit).
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/v0.64.0/src/qai_hub_models/models/efficientvit_b2_cls) 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 2.50, ONNX Runtime 1.30.0 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/efficientvit_b2_cls/releases/v0.64.0/efficientvit_b2_cls-onnx-float.zip)
| ONNX | w8a16 | Universal | QAIRT 2.50, ONNX Runtime 1.30.0 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/efficientvit_b2_cls/releases/v0.64.0/efficientvit_b2_cls-onnx-w8a16.zip)
| QNN_DLC | float | Universal | QAIRT 2.50 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/efficientvit_b2_cls/releases/v0.64.0/efficientvit_b2_cls-qnn_dlc-float.zip)
| QNN_DLC | w8a16 | Universal | QAIRT 2.50 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/efficientvit_b2_cls/releases/v0.64.0/efficientvit_b2_cls-qnn_dlc-w8a16.zip)
| TFLITE | float | Universal | QAIRT 2.50 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/efficientvit_b2_cls/releases/v0.64.0/efficientvit_b2_cls-tflite-float.zip)

For more device-specific assets and performance metrics, visit **[EfficientViT-b2-cls on Qualcomm® AI Hub](https://aihub.qualcomm.com/models/efficientvit_b2_cls)**.


### Option 2: Export with Custom Configurations

Use the [Qualcomm® AI Hub Models](https://github.com/qualcomm/ai-hub-models/blob/v0.64.0/src/qai_hub_models/models/efficientvit_b2_cls) 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 [EfficientViT-b2-cls on GitHub](https://github.com/qualcomm/ai-hub-models/blob/v0.64.0/src/qai_hub_models/models/efficientvit_b2_cls) for usage instructions.

## Model Details

**Model Type:** Model_use_case.image_classification

**Model Stats:**
- Input resolution: 224x224
- Model checkpoint: Imagenet
- Model size (float): 92.9 MB
- Number of parameters: 24.3M

## Performance Summary
| Model | Runtime | Precision | Chipset | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit
|---|---|---|---|---|---|---
| EfficientViT-b2-cls | ONNX | float | Snapdragon® 8 Elite Gen 5 For Galaxy Mobile | 2.433 ms | 1 - 76 MB | NPU
| EfficientViT-b2-cls | ONNX | float | Snapdragon® 8 Elite For Galaxy Mobile | 2.711 ms | 0 - 73 MB | NPU
| EfficientViT-b2-cls | ONNX | float | Snapdragon® X2 Elite | 2.524 ms | 2 - 2 MB | NPU
| EfficientViT-b2-cls | ONNX | float | Snapdragon® X Elite | 5.048 ms | 49 - 49 MB | NPU
| EfficientViT-b2-cls | ONNX | float | Snapdragon® 8 Gen 3 Mobile | 3.224 ms | 0 - 150 MB | NPU
| EfficientViT-b2-cls | ONNX | float | Snapdragon® 8 Gen 1 Mobile | 5.639 ms | 1 - 147 MB | NPU
| EfficientViT-b2-cls | ONNX | float | Qualcomm® Dragonwing™ IQ-8275 | 5.094 ms | 0 - 5 MB | NPU
| EfficientViT-b2-cls | ONNX | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 4.891 ms | 0 - 58 MB | NPU
| EfficientViT-b2-cls | ONNX | float | Qualcomm® QCS8450 | 5.639 ms | 1 - 147 MB | NPU
| EfficientViT-b2-cls | ONNX | float | Qualcomm® Dragonwing™ IQ-9075 | 5.595 ms | 1 - 4 MB | NPU
| EfficientViT-b2-cls | ONNX | float | Qualcomm® Dragonwing™ IQ-X7181 | 5.048 ms | 49 - 49 MB | NPU
| EfficientViT-b2-cls | ONNX | float | Qualcomm® Dragonwing™ Q-8750 | 2.711 ms | 0 - 73 MB | NPU
| EfficientViT-b2-cls | ONNX | w8a16 | Snapdragon® 8 Elite Gen 5 For Galaxy Mobile | 2.091 ms | 0 - 122 MB | NPU
| EfficientViT-b2-cls | ONNX | w8a16 | Snapdragon® 8 Elite For Galaxy Mobile | 2.433 ms | 0 - 120 MB | NPU
| EfficientViT-b2-cls | ONNX | w8a16 | Snapdragon® X2 Elite | 2.228 ms | 1 - 1 MB | NPU
| EfficientViT-b2-cls | ONNX | w8a16 | Snapdragon® X Elite | 4.671 ms | 26 - 26 MB | NPU
| EfficientViT-b2-cls | ONNX | w8a16 | Snapdragon® 8 Gen 3 Mobile | 3.004 ms | 0 - 163 MB | NPU
| EfficientViT-b2-cls | ONNX | w8a16 | Snapdragon® 8 Gen 1 Mobile | 5.333 ms | 0 - 176 MB | NPU
| EfficientViT-b2-cls | ONNX | w8a16 | Qualcomm® Dragonwing™ QCS6490 | 11.04 ms | 0 - 4 MB | NPU
| EfficientViT-b2-cls | ONNX | w8a16 | Qualcomm® Dragonwing™ IQ-8275 | 4.078 ms | 0 - 4 MB | NPU
| EfficientViT-b2-cls | ONNX | w8a16 | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 4.477 ms | 0 - 185 MB | NPU
| EfficientViT-b2-cls | ONNX | w8a16 | Qualcomm® QCS8450 | 5.333 ms | 0 - 176 MB | NPU
| EfficientViT-b2-cls | ONNX | w8a16 | Qualcomm® Dragonwing™ IQ-9075 | 4.762 ms | 0 - 4 MB | NPU
| EfficientViT-b2-cls | ONNX | w8a16 | Qualcomm® Dragonwing™ IQ-X7181 | 4.671 ms | 26 - 26 MB | NPU
| EfficientViT-b2-cls | ONNX | w8a16 | Qualcomm® Dragonwing™ Q-6690 | 17.239 ms | 0 - 239 MB | NPU
| EfficientViT-b2-cls | ONNX | w8a16 | Qualcomm® Dragonwing™ Q-7790 | 4.797 ms | 0 - 233 MB | NPU
| EfficientViT-b2-cls | ONNX | w8a16 | Qualcomm® Dragonwing™ Q-8750 | 2.433 ms | 0 - 120 MB | NPU
| EfficientViT-b2-cls | ONNX | w8a16 | Snapdragon® 7 Gen 4 Mobile | 4.797 ms | 0 - 233 MB | NPU
| EfficientViT-b2-cls | QNN_DLC | float | Snapdragon® 8 Elite Gen 5 For Galaxy Mobile | 2.077 ms | 1 - 72 MB | NPU
| EfficientViT-b2-cls | QNN_DLC | float | Snapdragon® 8 Elite For Galaxy Mobile | 2.368 ms | 1 - 70 MB | NPU
| EfficientViT-b2-cls | QNN_DLC | float | Snapdragon® X2 Elite | 2.49 ms | 1 - 1 MB | NPU
| EfficientViT-b2-cls | QNN_DLC | float | Snapdragon® X Elite | 4.853 ms | 1 - 1 MB | NPU
| EfficientViT-b2-cls | QNN_DLC | float | Snapdragon® 8 Gen 3 Mobile | 2.963 ms | 0 - 140 MB | NPU
| EfficientViT-b2-cls | QNN_DLC | float | Snapdragon® 8 Gen 1 Mobile | 5.878 ms | 0 - 145 MB | NPU
| EfficientViT-b2-cls | QNN_DLC | float | Qualcomm® Dragonwing™ IQ-8275 | 4.614 ms | 1 - 4 MB | NPU
| EfficientViT-b2-cls | QNN_DLC | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 4.502 ms | 1 - 3 MB | NPU
| EfficientViT-b2-cls | QNN_DLC | float | Qualcomm® SA8650P | 6.803 ms | 1 - 69 MB | NPU
| EfficientViT-b2-cls | QNN_DLC | float | Qualcomm® SA8255P | 6.803 ms | 1 - 69 MB | NPU
| EfficientViT-b2-cls | QNN_DLC | float | Qualcomm® QCS8450 | 5.878 ms | 0 - 145 MB | NPU
| EfficientViT-b2-cls | QNN_DLC | float | Qualcomm® Dragonwing™ IQ-9075 | 5.157 ms | 3 - 5 MB | NPU
| EfficientViT-b2-cls | QNN_DLC | float | Qualcomm® Dragonwing™ IQ-X7181 | 4.853 ms | 1 - 1 MB | NPU
| EfficientViT-b2-cls | QNN_DLC | float | Qualcomm® Dragonwing™ Q-8750 | 2.368 ms | 1 - 70 MB | NPU
| EfficientViT-b2-cls | QNN_DLC | float | Qualcomm® SA8295P | 5.928 ms | 0 - 71 MB | NPU
| EfficientViT-b2-cls | QNN_DLC | w8a16 | Snapdragon® 8 Elite Gen 5 For Galaxy Mobile | 2.098 ms | 0 - 113 MB | NPU
| EfficientViT-b2-cls | QNN_DLC | w8a16 | Snapdragon® 8 Elite For Galaxy Mobile | 2.445 ms | 0 - 111 MB | NPU
| EfficientViT-b2-cls | QNN_DLC | w8a16 | Snapdragon® X2 Elite | 2.572 ms | 0 - 0 MB | NPU
| EfficientViT-b2-cls | QNN_DLC | w8a16 | Snapdragon® X Elite | 5.471 ms | 0 - 0 MB | NPU
| EfficientViT-b2-cls | QNN_DLC | w8a16 | Snapdragon® 8 Gen 3 Mobile | 3.361 ms | 0 - 153 MB | NPU
| EfficientViT-b2-cls | QNN_DLC | w8a16 | Qualcomm® Dragonwing™ IQ-8275 | 4.503 ms | 0 - 4 MB | NPU
| EfficientViT-b2-cls | QNN_DLC | w8a16 | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 5.051 ms | 0 - 2 MB | NPU
| EfficientViT-b2-cls | QNN_DLC | w8a16 | Qualcomm® SA8650P | 5.456 ms | 0 - 106 MB | NPU
| EfficientViT-b2-cls | QNN_DLC | w8a16 | Qualcomm® SA8255P | 5.456 ms | 0 - 106 MB | NPU
| EfficientViT-b2-cls | QNN_DLC | w8a16 | Qualcomm® Dragonwing™ IQ-9075 | 5.337 ms | 0 - 3 MB | NPU
| EfficientViT-b2-cls | QNN_DLC | w8a16 | Qualcomm® Dragonwing™ IQ-X7181 | 5.471 ms | 0 - 0 MB | NPU
| EfficientViT-b2-cls | QNN_DLC | w8a16 | Qualcomm® Dragonwing™ Q-6690 | 24.119 ms | 0 - 225 MB | NPU
| EfficientViT-b2-cls | QNN_DLC | w8a16 | Qualcomm® Dragonwing™ Q-7790 | 5.326 ms | 0 - 218 MB | NPU
| EfficientViT-b2-cls | QNN_DLC | w8a16 | Qualcomm® Dragonwing™ Q-8750 | 2.445 ms | 0 - 111 MB | NPU
| EfficientViT-b2-cls | QNN_DLC | w8a16 | Snapdragon® 7 Gen 4 Mobile | 5.326 ms | 0 - 218 MB | NPU
| EfficientViT-b2-cls | TFLITE | float | Snapdragon® 8 Elite Gen 5 For Galaxy Mobile | 2.07 ms | 0 - 115 MB | NPU
| EfficientViT-b2-cls | TFLITE | float | Snapdragon® 8 Elite For Galaxy Mobile | 2.351 ms | 0 - 109 MB | NPU
| EfficientViT-b2-cls | TFLITE | float | Snapdragon® 8 Gen 3 Mobile | 2.946 ms | 0 - 179 MB | NPU
| EfficientViT-b2-cls | TFLITE | float | Snapdragon® 8 Gen 1 Mobile | 5.802 ms | 0 - 181 MB | NPU
| EfficientViT-b2-cls | TFLITE | float | Qualcomm® Dragonwing™ IQ-8275 | 4.652 ms | 0 - 53 MB | NPU
| EfficientViT-b2-cls | TFLITE | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 4.505 ms | 0 - 7 MB | NPU
| EfficientViT-b2-cls | TFLITE | float | Qualcomm® SA8650P | 6.819 ms | 0 - 111 MB | NPU
| EfficientViT-b2-cls | TFLITE | float | Qualcomm® SA8255P | 6.819 ms | 0 - 111 MB | NPU
| EfficientViT-b2-cls | TFLITE | float | Qualcomm® QCS8450 | 5.802 ms | 0 - 181 MB | NPU
| EfficientViT-b2-cls | TFLITE | float | Qualcomm® Dragonwing™ IQ-9075 | 5.117 ms | 0 - 52 MB | NPU
| EfficientViT-b2-cls | TFLITE | float | Qualcomm® Dragonwing™ Q-8750 | 2.351 ms | 0 - 109 MB | NPU
| EfficientViT-b2-cls | TFLITE | float | Qualcomm® SA8295P | 5.924 ms | 0 - 108 MB | NPU

## License
* The license for the original implementation of EfficientViT-b2-cls can be found
  [here](https://github.com/CVHub520/efficientvit/blob/main/LICENSE).

## References
* [EfficientViT: Multi-Scale Linear Attention for High-Resolution Dense Prediction](https://arxiv.org/abs/2205.14756)
* [Source Model Implementation](https://github.com/CVHub520/efficientvit)

## 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).