README.md

---
library_name: pytorch
license: other
tags:
- android
pipeline_tag: image-classification

---

![](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/shufflenet_v2/web-assets/model_demo.png)

# Shufflenet-v2: Optimized for Mobile Deployment
## Imagenet classifier and general purpose backbone


ShufflenetV2 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 model is an implementation of Shufflenet-v2 found [here](https://github.com/pytorch/vision/blob/main/torchvision/models/shufflenetv2.py).


This repository provides scripts to run Shufflenet-v2 on Qualcomm® devices.
More details on model performance across various devices, can be found
[here](https://aihub.qualcomm.com/models/shufflenet_v2).


### Model Details

- **Model Type:** Model_use_case.image_classification
- **Model Stats:**
  - Model checkpoint: Imagenet
  - Input resolution: 224x224
  - Number of parameters: 1.36M
  - Model size (float): 5.25 MB
  - Model size (w8a8): 4.42 MB

| Model | Precision | Device | Chipset | Target Runtime | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit | Target Model
|---|---|---|---|---|---|---|---|---|
| Shufflenet-v2 | float | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | TFLITE | 1.627 ms | 0 - 15 MB | NPU | [Shufflenet-v2.tflite](https://huggingface.co/qualcomm/Shufflenet-v2/blob/main/Shufflenet-v2.tflite) |
| Shufflenet-v2 | float | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | QNN | 1.598 ms | 1 - 11 MB | NPU | Use Export Script |
| Shufflenet-v2 | float | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | TFLITE | 0.869 ms | 0 - 30 MB | NPU | [Shufflenet-v2.tflite](https://huggingface.co/qualcomm/Shufflenet-v2/blob/main/Shufflenet-v2.tflite) |
| Shufflenet-v2 | float | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | QNN | 1.303 ms | 1 - 23 MB | NPU | Use Export Script |
| Shufflenet-v2 | float | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | TFLITE | 0.694 ms | 0 - 27 MB | NPU | [Shufflenet-v2.tflite](https://huggingface.co/qualcomm/Shufflenet-v2/blob/main/Shufflenet-v2.tflite) |
| Shufflenet-v2 | float | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | QNN | 0.692 ms | 1 - 3 MB | NPU | Use Export Script |
| Shufflenet-v2 | float | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | TFLITE | 0.97 ms | 0 - 18 MB | NPU | [Shufflenet-v2.tflite](https://huggingface.co/qualcomm/Shufflenet-v2/blob/main/Shufflenet-v2.tflite) |
| Shufflenet-v2 | float | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | QNN | 0.923 ms | 1 - 15 MB | NPU | Use Export Script |
| Shufflenet-v2 | float | SA7255P ADP | Qualcomm® SA7255P | TFLITE | 1.627 ms | 0 - 15 MB | NPU | [Shufflenet-v2.tflite](https://huggingface.co/qualcomm/Shufflenet-v2/blob/main/Shufflenet-v2.tflite) |
| Shufflenet-v2 | float | SA7255P ADP | Qualcomm® SA7255P | QNN | 1.598 ms | 1 - 11 MB | NPU | Use Export Script |
| Shufflenet-v2 | float | SA8255 (Proxy) | Qualcomm® SA8255P (Proxy) | TFLITE | 0.692 ms | 0 - 26 MB | NPU | [Shufflenet-v2.tflite](https://huggingface.co/qualcomm/Shufflenet-v2/blob/main/Shufflenet-v2.tflite) |
| Shufflenet-v2 | float | SA8255 (Proxy) | Qualcomm® SA8255P (Proxy) | QNN | 0.69 ms | 1 - 3 MB | NPU | Use Export Script |
| Shufflenet-v2 | float | SA8295P ADP | Qualcomm® SA8295P | TFLITE | 1.232 ms | 0 - 19 MB | NPU | [Shufflenet-v2.tflite](https://huggingface.co/qualcomm/Shufflenet-v2/blob/main/Shufflenet-v2.tflite) |
| Shufflenet-v2 | float | SA8295P ADP | Qualcomm® SA8295P | QNN | 1.183 ms | 0 - 18 MB | NPU | Use Export Script |
| Shufflenet-v2 | float | SA8650 (Proxy) | Qualcomm® SA8650P (Proxy) | TFLITE | 0.692 ms | 0 - 27 MB | NPU | [Shufflenet-v2.tflite](https://huggingface.co/qualcomm/Shufflenet-v2/blob/main/Shufflenet-v2.tflite) |
| Shufflenet-v2 | float | SA8650 (Proxy) | Qualcomm® SA8650P (Proxy) | QNN | 0.696 ms | 1 - 3 MB | NPU | Use Export Script |
| Shufflenet-v2 | float | SA8775P ADP | Qualcomm® SA8775P | TFLITE | 0.97 ms | 0 - 18 MB | NPU | [Shufflenet-v2.tflite](https://huggingface.co/qualcomm/Shufflenet-v2/blob/main/Shufflenet-v2.tflite) |
| Shufflenet-v2 | float | SA8775P ADP | Qualcomm® SA8775P | QNN | 0.923 ms | 1 - 15 MB | NPU | Use Export Script |
| Shufflenet-v2 | float | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 Mobile | TFLITE | 0.697 ms | 0 - 28 MB | NPU | [Shufflenet-v2.tflite](https://huggingface.co/qualcomm/Shufflenet-v2/blob/main/Shufflenet-v2.tflite) |
| Shufflenet-v2 | float | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 Mobile | QNN | 0.689 ms | 0 - 17 MB | NPU | Use Export Script |
| Shufflenet-v2 | float | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 Mobile | ONNX | 0.973 ms | 0 - 15 MB | NPU | [Shufflenet-v2.onnx](https://huggingface.co/qualcomm/Shufflenet-v2/blob/main/Shufflenet-v2.onnx) |
| Shufflenet-v2 | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | TFLITE | 0.451 ms | 0 - 28 MB | NPU | [Shufflenet-v2.tflite](https://huggingface.co/qualcomm/Shufflenet-v2/blob/main/Shufflenet-v2.tflite) |
| Shufflenet-v2 | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | QNN | 0.464 ms | 0 - 24 MB | NPU | Use Export Script |
| Shufflenet-v2 | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | ONNX | 0.595 ms | 0 - 26 MB | NPU | [Shufflenet-v2.onnx](https://huggingface.co/qualcomm/Shufflenet-v2/blob/main/Shufflenet-v2.onnx) |
| Shufflenet-v2 | float | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite Mobile | TFLITE | 0.444 ms | 0 - 21 MB | NPU | [Shufflenet-v2.tflite](https://huggingface.co/qualcomm/Shufflenet-v2/blob/main/Shufflenet-v2.tflite) |
| Shufflenet-v2 | float | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite Mobile | QNN | 0.427 ms | 0 - 17 MB | NPU | Use Export Script |
| Shufflenet-v2 | float | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite Mobile | ONNX | 0.655 ms | 1 - 22 MB | NPU | [Shufflenet-v2.onnx](https://huggingface.co/qualcomm/Shufflenet-v2/blob/main/Shufflenet-v2.onnx) |
| Shufflenet-v2 | float | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN | 0.823 ms | 1 - 1 MB | NPU | Use Export Script |
| Shufflenet-v2 | float | Snapdragon X Elite CRD | Snapdragon® X Elite | ONNX | 1.026 ms | 0 - 0 MB | NPU | [Shufflenet-v2.onnx](https://huggingface.co/qualcomm/Shufflenet-v2/blob/main/Shufflenet-v2.onnx) |
| Shufflenet-v2 | w8a8 | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | TFLITE | 0.797 ms | 0 - 12 MB | NPU | [Shufflenet-v2.tflite](https://huggingface.co/qualcomm/Shufflenet-v2/blob/main/Shufflenet-v2_w8a8.tflite) |
| Shufflenet-v2 | w8a8 | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | QNN | 1.08 ms | 0 - 10 MB | NPU | Use Export Script |
| Shufflenet-v2 | w8a8 | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | TFLITE | 0.424 ms | 0 - 24 MB | NPU | [Shufflenet-v2.tflite](https://huggingface.co/qualcomm/Shufflenet-v2/blob/main/Shufflenet-v2_w8a8.tflite) |
| Shufflenet-v2 | w8a8 | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | QNN | 0.623 ms | 0 - 19 MB | NPU | Use Export Script |
| Shufflenet-v2 | w8a8 | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | TFLITE | 0.327 ms | 0 - 9 MB | NPU | [Shufflenet-v2.tflite](https://huggingface.co/qualcomm/Shufflenet-v2/blob/main/Shufflenet-v2_w8a8.tflite) |
| Shufflenet-v2 | w8a8 | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | QNN | 0.475 ms | 0 - 2 MB | NPU | Use Export Script |
| Shufflenet-v2 | w8a8 | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | TFLITE | 0.522 ms | 0 - 14 MB | NPU | [Shufflenet-v2.tflite](https://huggingface.co/qualcomm/Shufflenet-v2/blob/main/Shufflenet-v2_w8a8.tflite) |
| Shufflenet-v2 | w8a8 | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | QNN | 0.674 ms | 0 - 15 MB | NPU | Use Export Script |
| Shufflenet-v2 | w8a8 | RB3 Gen 2 (Proxy) | Qualcomm® QCS6490 (Proxy) | TFLITE | 0.645 ms | 0 - 16 MB | NPU | [Shufflenet-v2.tflite](https://huggingface.co/qualcomm/Shufflenet-v2/blob/main/Shufflenet-v2_w8a8.tflite) |
| Shufflenet-v2 | w8a8 | RB3 Gen 2 (Proxy) | Qualcomm® QCS6490 (Proxy) | QNN | 1.066 ms | 0 - 12 MB | NPU | Use Export Script |
| Shufflenet-v2 | w8a8 | RB5 (Proxy) | Qualcomm® QCS8250 (Proxy) | TFLITE | 7.248 ms | 0 - 8 MB | CPU | [Shufflenet-v2.tflite](https://huggingface.co/qualcomm/Shufflenet-v2/blob/main/Shufflenet-v2_w8a8.tflite) |
| Shufflenet-v2 | w8a8 | SA7255P ADP | Qualcomm® SA7255P | TFLITE | 0.797 ms | 0 - 12 MB | NPU | [Shufflenet-v2.tflite](https://huggingface.co/qualcomm/Shufflenet-v2/blob/main/Shufflenet-v2_w8a8.tflite) |
| Shufflenet-v2 | w8a8 | SA7255P ADP | Qualcomm® SA7255P | QNN | 1.08 ms | 0 - 10 MB | NPU | Use Export Script |
| Shufflenet-v2 | w8a8 | SA8255 (Proxy) | Qualcomm® SA8255P (Proxy) | TFLITE | 0.325 ms | 0 - 8 MB | NPU | [Shufflenet-v2.tflite](https://huggingface.co/qualcomm/Shufflenet-v2/blob/main/Shufflenet-v2_w8a8.tflite) |
| Shufflenet-v2 | w8a8 | SA8255 (Proxy) | Qualcomm® SA8255P (Proxy) | QNN | 0.473 ms | 0 - 2 MB | NPU | Use Export Script |
| Shufflenet-v2 | w8a8 | SA8295P ADP | Qualcomm® SA8295P | TFLITE | 0.647 ms | 0 - 17 MB | NPU | [Shufflenet-v2.tflite](https://huggingface.co/qualcomm/Shufflenet-v2/blob/main/Shufflenet-v2_w8a8.tflite) |
| Shufflenet-v2 | w8a8 | SA8295P ADP | Qualcomm® SA8295P | QNN | 0.815 ms | 0 - 18 MB | NPU | Use Export Script |
| Shufflenet-v2 | w8a8 | SA8650 (Proxy) | Qualcomm® SA8650P (Proxy) | TFLITE | 0.322 ms | 0 - 9 MB | NPU | [Shufflenet-v2.tflite](https://huggingface.co/qualcomm/Shufflenet-v2/blob/main/Shufflenet-v2_w8a8.tflite) |
| Shufflenet-v2 | w8a8 | SA8650 (Proxy) | Qualcomm® SA8650P (Proxy) | QNN | 0.469 ms | 0 - 2 MB | NPU | Use Export Script |
| Shufflenet-v2 | w8a8 | SA8775P ADP | Qualcomm® SA8775P | TFLITE | 0.522 ms | 0 - 14 MB | NPU | [Shufflenet-v2.tflite](https://huggingface.co/qualcomm/Shufflenet-v2/blob/main/Shufflenet-v2_w8a8.tflite) |
| Shufflenet-v2 | w8a8 | SA8775P ADP | Qualcomm® SA8775P | QNN | 0.674 ms | 0 - 15 MB | NPU | Use Export Script |
| Shufflenet-v2 | w8a8 | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 Mobile | TFLITE | 0.326 ms | 0 - 9 MB | NPU | [Shufflenet-v2.tflite](https://huggingface.co/qualcomm/Shufflenet-v2/blob/main/Shufflenet-v2_w8a8.tflite) |
| Shufflenet-v2 | w8a8 | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 Mobile | QNN | 0.471 ms | 0 - 9 MB | NPU | Use Export Script |
| Shufflenet-v2 | w8a8 | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 Mobile | ONNX | 0.794 ms | 0 - 8 MB | NPU | [Shufflenet-v2.onnx](https://huggingface.co/qualcomm/Shufflenet-v2/blob/main/Shufflenet-v2_w8a8.onnx) |
| Shufflenet-v2 | w8a8 | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | TFLITE | 0.226 ms | 0 - 25 MB | NPU | [Shufflenet-v2.tflite](https://huggingface.co/qualcomm/Shufflenet-v2/blob/main/Shufflenet-v2_w8a8.tflite) |
| Shufflenet-v2 | w8a8 | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | QNN | 0.34 ms | 0 - 25 MB | NPU | Use Export Script |
| Shufflenet-v2 | w8a8 | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | ONNX | 0.532 ms | 0 - 26 MB | NPU | [Shufflenet-v2.onnx](https://huggingface.co/qualcomm/Shufflenet-v2/blob/main/Shufflenet-v2_w8a8.onnx) |
| Shufflenet-v2 | w8a8 | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite Mobile | TFLITE | 0.255 ms | 0 - 19 MB | NPU | [Shufflenet-v2.tflite](https://huggingface.co/qualcomm/Shufflenet-v2/blob/main/Shufflenet-v2_w8a8.tflite) |
| Shufflenet-v2 | w8a8 | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite Mobile | QNN | 0.316 ms | 0 - 18 MB | NPU | Use Export Script |
| Shufflenet-v2 | w8a8 | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite Mobile | ONNX | 0.597 ms | 0 - 21 MB | NPU | [Shufflenet-v2.onnx](https://huggingface.co/qualcomm/Shufflenet-v2/blob/main/Shufflenet-v2_w8a8.onnx) |
| Shufflenet-v2 | w8a8 | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN | 0.567 ms | 0 - 0 MB | NPU | Use Export Script |
| Shufflenet-v2 | w8a8 | Snapdragon X Elite CRD | Snapdragon® X Elite | ONNX | 0.781 ms | 0 - 0 MB | NPU | [Shufflenet-v2.onnx](https://huggingface.co/qualcomm/Shufflenet-v2/blob/main/Shufflenet-v2_w8a8.onnx) |




## Installation


Install the package via pip:
```bash
pip install qai-hub-models
```


## Configure Qualcomm® AI Hub to run this model on a cloud-hosted device

Sign-in to [Qualcomm® AI Hub](https://app.aihub.qualcomm.com/) with your
Qualcomm® ID. Once signed in navigate to `Account -> Settings -> API Token`.

With this API token, you can configure your client to run models on the cloud
hosted devices.
```bash
qai-hub configure --api_token API_TOKEN
```
Navigate to [docs](https://app.aihub.qualcomm.com/docs/) for more information.



## Demo off target

The package contains a simple end-to-end demo that downloads pre-trained
weights and runs this model on a sample input.

```bash
python -m qai_hub_models.models.shufflenet_v2.demo
```

The above demo runs a reference implementation of pre-processing, model
inference, and post processing.

**NOTE**: If you want running in a Jupyter Notebook or Google Colab like
environment, please add the following to your cell (instead of the above).
```
%run -m qai_hub_models.models.shufflenet_v2.demo
```


### Run model on a cloud-hosted device

In addition to the demo, you can also run the model on a cloud-hosted Qualcomm®
device. This script does the following:
* Performance check on-device on a cloud-hosted device
* Downloads compiled assets that can be deployed on-device for Android.
* Accuracy check between PyTorch and on-device outputs.

```bash
python -m qai_hub_models.models.shufflenet_v2.export
```
```
Profiling Results
------------------------------------------------------------
Shufflenet-v2
Device                          : cs_8275 (ANDROID 14)                 
Runtime                         : TFLITE                               
Estimated inference time (ms)   : 1.6                                  
Estimated peak memory usage (MB): [0, 15]                              
Total # Ops                     : 159                                  
Compute Unit(s)                 : npu (159 ops) gpu (0 ops) cpu (0 ops)
```


## How does this work?

This [export script](https://aihub.qualcomm.com/models/shufflenet_v2/qai_hub_models/models/Shufflenet-v2/export.py)
leverages [Qualcomm® AI Hub](https://aihub.qualcomm.com/) to optimize, validate, and deploy this model
on-device. Lets go through each step below in detail:

Step 1: **Compile model for on-device deployment**

To compile a PyTorch model for on-device deployment, we first trace the model
in memory using the `jit.trace` and then call the `submit_compile_job` API.

```python
import torch

import qai_hub as hub
from qai_hub_models.models.shufflenet_v2 import Model

# Load the model
torch_model = Model.from_pretrained()

# Device
device = hub.Device("Samsung Galaxy S24")

# Trace model
input_shape = torch_model.get_input_spec()
sample_inputs = torch_model.sample_inputs()

pt_model = torch.jit.trace(torch_model, [torch.tensor(data[0]) for _, data in sample_inputs.items()])

# Compile model on a specific device
compile_job = hub.submit_compile_job(
    model=pt_model,
    device=device,
    input_specs=torch_model.get_input_spec(),
)

# Get target model to run on-device
target_model = compile_job.get_target_model()

```


Step 2: **Performance profiling on cloud-hosted device**

After compiling models from step 1. Models can be profiled model on-device using the
`target_model`. Note that this scripts runs the model on a device automatically
provisioned in the cloud.  Once the job is submitted, you can navigate to a
provided job URL to view a variety of on-device performance metrics.
```python
profile_job = hub.submit_profile_job(
    model=target_model,
    device=device,
)
        
```

Step 3: **Verify on-device accuracy**

To verify the accuracy of the model on-device, you can run on-device inference
on sample input data on the same cloud hosted device.
```python
input_data = torch_model.sample_inputs()
inference_job = hub.submit_inference_job(
    model=target_model,
    device=device,
    inputs=input_data,
)
    on_device_output = inference_job.download_output_data()

```
With the output of the model, you can compute like PSNR, relative errors or
spot check the output with expected output.

**Note**: This on-device profiling and inference requires access to Qualcomm®
AI Hub. [Sign up for access](https://myaccount.qualcomm.com/signup).



## Run demo on a cloud-hosted device

You can also run the demo on-device.

```bash
python -m qai_hub_models.models.shufflenet_v2.demo --on-device
```

**NOTE**: If you want running in a Jupyter Notebook or Google Colab like
environment, please add the following to your cell (instead of the above).
```
%run -m qai_hub_models.models.shufflenet_v2.demo -- --on-device
```


## Deploying compiled model to Android


The models can be deployed using multiple runtimes:
- TensorFlow Lite (`.tflite` export): [This
  tutorial](https://www.tensorflow.org/lite/android/quickstart) provides a
  guide to deploy the .tflite model in an Android application.


- QNN (`.so` export ): This [sample
  app](https://docs.qualcomm.com/bundle/publicresource/topics/80-63442-50/sample_app.html)
provides instructions on how to use the `.so` shared library  in an Android application.


## View on Qualcomm® AI Hub
Get more details on Shufflenet-v2's performance across various devices [here](https://aihub.qualcomm.com/models/shufflenet_v2).
Explore all available models on [Qualcomm® AI Hub](https://aihub.qualcomm.com/)


## License
* The license for the original implementation of Shufflenet-v2 can be found
  [here](https://github.com/pytorch/vision/blob/main/LICENSE).
* The license for the compiled assets for on-device deployment can be found [here](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/Qualcomm+AI+Hub+Proprietary+License.pdf)



## References
* [ShuffleNet V2: Practical Guidelines for Efficient CNN Architecture Design](https://arxiv.org/abs/1807.11164)
* [Source Model Implementation](https://github.com/pytorch/vision/blob/main/torchvision/models/shufflenetv2.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).