README.md

---
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/mnasnet05/web-assets/model_demo.png)

# MNASNet05: Optimized for Qualcomm Devices

MNASNet05 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 MNASNet05 found [here](https://github.com/pytorch/vision/blob/main/torchvision/models/mnasnet.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/v0.59.0/src/qai_hub_models/models/mnasnet05) 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 2.45, ONNX Runtime 1.27.1 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/mnasnet05/releases/v0.59.0/mnasnet05-onnx-float.zip)
| ONNX | w8a16 | Universal | QAIRT 2.45, ONNX Runtime 1.27.1 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/mnasnet05/releases/v0.59.0/mnasnet05-onnx-w8a16.zip)
| QNN_DLC | float | Universal | QAIRT 2.45 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/mnasnet05/releases/v0.59.0/mnasnet05-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/mnasnet05/releases/v0.59.0/mnasnet05-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/mnasnet05/releases/v0.59.0/mnasnet05-tflite-float.zip)

For more device-specific assets and performance metrics, visit **[MNASNet05 on Qualcomm® AI Hub](https://aihub.qualcomm.com/models/mnasnet05)**.


### Option 2: Export with Custom Configurations

Use the [Qualcomm® AI Hub Models](https://github.com/qualcomm/ai-hub-models/blob/v0.59.0/src/qai_hub_models/models/mnasnet05) 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 [MNASNet05 on GitHub](https://github.com/qualcomm/ai-hub-models/blob/v0.59.0/src/qai_hub_models/models/mnasnet05) for usage instructions.

## Model Details

**Model Type:** Model_use_case.image_classification

**Model Stats:**
- Model checkpoint: Imagenet
- Input resolution: 224x224
- Number of parameters: 2.21M
- Model size (float): 8.45 MB
- Model size (w8a16): 2.79 MB

## Performance Summary
| Model | Runtime | Precision | Chipset | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit
|---|---|---|---|---|---|---
| MNASNet05 | ONNX | float | Snapdragon® X2 Elite | 0.229 ms | 2 - 2 MB | NPU
| MNASNet05 | ONNX | float | Snapdragon® X Elite | 0.493 ms | 5 - 5 MB | NPU
| MNASNet05 | ONNX | float | Snapdragon® 8 Gen 3 Mobile | 0.328 ms | 0 - 47 MB | NPU
| MNASNet05 | ONNX | float | Snapdragon® 8 Gen 1 Mobile | 0.873 ms | 1 - 52 MB | NPU
| MNASNet05 | ONNX | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 0.49 ms | 0 - 7 MB | NPU
| MNASNet05 | ONNX | float | Qualcomm® QCS8450 | 0.873 ms | 1 - 52 MB | NPU
| MNASNet05 | ONNX | float | Qualcomm® Dragonwing™ IQ-9075 | 0.751 ms | 1 - 3 MB | NPU
| MNASNet05 | ONNX | float | Qualcomm® Dragonwing™ IQ-X7181 | 0.493 ms | 5 - 5 MB | NPU
| MNASNet05 | ONNX | float | Qualcomm® Dragonwing™ Q-8750 | 0.266 ms | 0 - 31 MB | NPU
| MNASNet05 | ONNX | float | Snapdragon® 8 Elite Mobile | 0.266 ms | 0 - 31 MB | NPU
| MNASNet05 | ONNX | float | Snapdragon® 8 Elite Gen 5 Mobile | 0.232 ms | 0 - 28 MB | NPU
| MNASNet05 | ONNX | w8a16 | Snapdragon® X2 Elite | 0.22 ms | 1 - 1 MB | NPU
| MNASNet05 | ONNX | w8a16 | Snapdragon® X Elite | 0.516 ms | 0 - 0 MB | NPU
| MNASNet05 | ONNX | w8a16 | Snapdragon® 8 Gen 3 Mobile | 0.341 ms | 0 - 43 MB | NPU
| MNASNet05 | ONNX | w8a16 | Snapdragon® 8 Gen 1 Mobile | 0.676 ms | 0 - 48 MB | NPU
| MNASNet05 | ONNX | w8a16 | Qualcomm® Dragonwing™ QCS6490 | 2.134 ms | 0 - 3 MB | NPU
| MNASNet05 | ONNX | w8a16 | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 0.511 ms | 0 - 5 MB | NPU
| MNASNet05 | ONNX | w8a16 | Qualcomm® QCS8450 | 0.676 ms | 0 - 48 MB | NPU
| MNASNet05 | ONNX | w8a16 | Qualcomm® Dragonwing™ IQ-9075 | 0.656 ms | 0 - 3 MB | NPU
| MNASNet05 | ONNX | w8a16 | Qualcomm® Dragonwing™ IQ-X7181 | 0.516 ms | 0 - 0 MB | NPU
| MNASNet05 | ONNX | w8a16 | Qualcomm® Dragonwing™ Q-8750 | 0.26 ms | 0 - 35 MB | NPU
| MNASNet05 | ONNX | w8a16 | Snapdragon® 8 Elite Mobile | 0.26 ms | 0 - 35 MB | NPU
| MNASNet05 | ONNX | w8a16 | Snapdragon® 8 Elite Gen 5 Mobile | 0.221 ms | 0 - 31 MB | NPU
| MNASNet05 | QNN_DLC | float | Snapdragon® X2 Elite | 0.417 ms | 1 - 1 MB | NPU
| MNASNet05 | QNN_DLC | float | Snapdragon® X Elite | 0.919 ms | 1 - 1 MB | NPU
| MNASNet05 | QNN_DLC | float | Snapdragon® 8 Gen 3 Mobile | 0.509 ms | 0 - 45 MB | NPU
| MNASNet05 | QNN_DLC | float | Snapdragon® 8 Gen 1 Mobile | 1.562 ms | 0 - 52 MB | NPU
| MNASNet05 | QNN_DLC | float | Qualcomm® Dragonwing™ QCS8275 | 2.301 ms | 1 - 28 MB | NPU
| MNASNet05 | QNN_DLC | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 0.788 ms | 1 - 7 MB | NPU
| MNASNet05 | QNN_DLC | float | Qualcomm® SA8775P | 1.102 ms | 0 - 30 MB | NPU
| MNASNet05 | QNN_DLC | float | Qualcomm® SA8650P | 1.102 ms | 0 - 30 MB | NPU
| MNASNet05 | QNN_DLC | float | Qualcomm® SA8255P | 1.102 ms | 0 - 30 MB | NPU
| MNASNet05 | QNN_DLC | float | Qualcomm® QCS8450 | 1.562 ms | 0 - 52 MB | NPU
| MNASNet05 | QNN_DLC | float | Qualcomm® Dragonwing™ IQ-9075 | 0.981 ms | 1 - 3 MB | NPU
| MNASNet05 | QNN_DLC | float | Qualcomm® Dragonwing™ IQ-X7181 | 0.919 ms | 1 - 1 MB | NPU
| MNASNet05 | QNN_DLC | float | Qualcomm® Dragonwing™ Q-8750 | 0.385 ms | 1 - 34 MB | NPU
| MNASNet05 | QNN_DLC | float | Qualcomm® SA7255P | 2.301 ms | 1 - 28 MB | NPU
| MNASNet05 | QNN_DLC | float | Qualcomm® SA8295P | 1.42 ms | 0 - 28 MB | NPU
| MNASNet05 | QNN_DLC | float | Snapdragon® 8 Elite Mobile | 0.385 ms | 1 - 34 MB | NPU
| MNASNet05 | QNN_DLC | float | Snapdragon® 8 Elite Gen 5 Mobile | 0.293 ms | 0 - 29 MB | NPU
| MNASNet05 | QNN_DLC | w8a16 | Snapdragon® X2 Elite | 0.396 ms | 0 - 0 MB | NPU
| MNASNet05 | QNN_DLC | w8a16 | Snapdragon® X Elite | 0.89 ms | 0 - 0 MB | NPU
| MNASNet05 | QNN_DLC | w8a16 | Snapdragon® 8 Gen 3 Mobile | 0.529 ms | 0 - 39 MB | NPU
| MNASNet05 | QNN_DLC | w8a16 | Snapdragon® 8 Gen 1 Mobile | 0.954 ms | 0 - 44 MB | NPU
| MNASNet05 | QNN_DLC | w8a16 | Qualcomm® Dragonwing™ QCS6490 | 2.721 ms | 0 - 2 MB | NPU
| MNASNet05 | QNN_DLC | w8a16 | Qualcomm® Dragonwing™ QCS8275 | 1.654 ms | 0 - 28 MB | NPU
| MNASNet05 | QNN_DLC | w8a16 | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 0.767 ms | 0 - 2 MB | NPU
| MNASNet05 | QNN_DLC | w8a16 | Qualcomm® SA8775P | 0.956 ms | 0 - 30 MB | NPU
| MNASNet05 | QNN_DLC | w8a16 | Qualcomm® SA8650P | 0.956 ms | 0 - 30 MB | NPU
| MNASNet05 | QNN_DLC | w8a16 | Qualcomm® SA8255P | 0.956 ms | 0 - 30 MB | NPU
| MNASNet05 | QNN_DLC | w8a16 | Qualcomm® QCS8450 | 0.954 ms | 0 - 44 MB | NPU
| MNASNet05 | QNN_DLC | w8a16 | Qualcomm® Dragonwing™ IQ-9075 | 0.867 ms | 0 - 2 MB | NPU
| MNASNet05 | QNN_DLC | w8a16 | Qualcomm® Dragonwing™ IQ-X7181 | 0.89 ms | 0 - 0 MB | NPU
| MNASNet05 | QNN_DLC | w8a16 | Qualcomm® Dragonwing™ Q-6690 | 3.05 ms | 0 - 140 MB | NPU
| MNASNet05 | QNN_DLC | w8a16 | Qualcomm® Dragonwing™ Q-7790 | 0.789 ms | 0 - 27 MB | NPU
| MNASNet05 | QNN_DLC | w8a16 | Qualcomm® Dragonwing™ Q-8750 | 0.353 ms | 0 - 27 MB | NPU
| MNASNet05 | QNN_DLC | w8a16 | Qualcomm® SA7255P | 1.654 ms | 0 - 28 MB | NPU
| MNASNet05 | QNN_DLC | w8a16 | Qualcomm® SA8295P | 1.224 ms | 0 - 26 MB | NPU
| MNASNet05 | QNN_DLC | w8a16 | Snapdragon® 8 Elite Mobile | 0.353 ms | 0 - 27 MB | NPU
| MNASNet05 | QNN_DLC | w8a16 | Snapdragon® 8 Elite Gen 5 Mobile | 0.304 ms | 0 - 28 MB | NPU
| MNASNet05 | QNN_DLC | w8a16 | Snapdragon® 7 Gen 4 Mobile | 0.789 ms | 0 - 27 MB | NPU
| MNASNet05 | TFLITE | float | Snapdragon® 8 Gen 3 Mobile | 0.526 ms | 0 - 45 MB | NPU
| MNASNet05 | TFLITE | float | Snapdragon® 8 Gen 1 Mobile | 1.573 ms | 0 - 53 MB | NPU
| MNASNet05 | TFLITE | float | Qualcomm® Dragonwing™ QCS8275 | 2.332 ms | 0 - 28 MB | NPU
| MNASNet05 | TFLITE | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 0.8 ms | 0 - 1 MB | NPU
| MNASNet05 | TFLITE | float | Qualcomm® SA8775P | 1.124 ms | 0 - 31 MB | NPU
| MNASNet05 | TFLITE | float | Qualcomm® SA8650P | 1.124 ms | 0 - 31 MB | NPU
| MNASNet05 | TFLITE | float | Qualcomm® SA8255P | 1.124 ms | 0 - 31 MB | NPU
| MNASNet05 | TFLITE | float | Qualcomm® QCS8450 | 1.573 ms | 0 - 53 MB | NPU
| MNASNet05 | TFLITE | float | Qualcomm® Dragonwing™ IQ-9075 | 0.986 ms | 0 - 8 MB | NPU
| MNASNet05 | TFLITE | float | Qualcomm® Dragonwing™ Q-8750 | 0.382 ms | 0 - 32 MB | NPU
| MNASNet05 | TFLITE | float | Qualcomm® SA7255P | 2.332 ms | 0 - 28 MB | NPU
| MNASNet05 | TFLITE | float | Qualcomm® SA8295P | 1.446 ms | 0 - 28 MB | NPU
| MNASNet05 | TFLITE | float | Snapdragon® 8 Elite Mobile | 0.382 ms | 0 - 32 MB | NPU
| MNASNet05 | TFLITE | float | Snapdragon® 8 Elite Gen 5 Mobile | 0.299 ms | 0 - 30 MB | NPU

## License
* The license for the original implementation of MNASNet05 can be found
  [here](https://github.com/pytorch/vision/blob/main/LICENSE).

## References
* [MnasNet: Platform-Aware Neural Architecture Search for Mobile](https://arxiv.org/abs/1807.11626)
* [Source Model Implementation](https://github.com/pytorch/vision/blob/main/torchvision/models/mnasnet.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).