Swin-Small: Optimized for Mobile Deployment

Imagenet classifier and general purpose backbone

SwinSmall 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 Swin-Small found here.

This repository provides scripts to run Swin-Small on Qualcomm® devices. More details on model performance across various devices, can be found here.

Model Details

• Model Type: Model_use_case.image_classification
• Model Stats:
  • Model checkpoint: Imagenet
  • Input resolution: 224x224
  • Number of parameters: 50.4M
  • Model size (float): 193 MB
  • Model size (w8a16): 52.5 MB

Model	Precision	Device	Chipset	Target Runtime	Inference Time (ms)	Peak Memory Range (MB)	Primary Compute Unit	Target Model
Swin-Small	float	QCS8275 (Proxy)	Qualcomm® QCS8275 (Proxy)	TFLITE	41.037 ms	0 - 425 MB	NPU	Swin-Small.tflite
Swin-Small	float	QCS8450 (Proxy)	Qualcomm® QCS8450 (Proxy)	TFLITE	25.56 ms	0 - 524 MB	NPU	Swin-Small.tflite
Swin-Small	float	QCS8550 (Proxy)	Qualcomm® QCS8550 (Proxy)	TFLITE	17.621 ms	0 - 5 MB	NPU	Swin-Small.tflite
Swin-Small	float	QCS8550 (Proxy)	Qualcomm® QCS8550 (Proxy)	ONNX	15.658 ms	0 - 114 MB	NPU	Swin-Small.onnx.zip
Swin-Small	float	QCS9075 (Proxy)	Qualcomm® QCS9075 (Proxy)	TFLITE	19.695 ms	0 - 684 MB	NPU	Swin-Small.tflite
Swin-Small	float	SA7255P ADP	Qualcomm® SA7255P	TFLITE	41.037 ms	0 - 425 MB	NPU	Swin-Small.tflite
Swin-Small	float	SA8255 (Proxy)	Qualcomm® SA8255P (Proxy)	TFLITE	17.597 ms	0 - 3 MB	NPU	Swin-Small.tflite
Swin-Small	float	SA8295P ADP	Qualcomm® SA8295P	TFLITE	25.348 ms	0 - 415 MB	NPU	Swin-Small.tflite
Swin-Small	float	SA8650 (Proxy)	Qualcomm® SA8650P (Proxy)	TFLITE	17.55 ms	0 - 4 MB	NPU	Swin-Small.tflite
Swin-Small	float	SA8775P ADP	Qualcomm® SA8775P	TFLITE	19.695 ms	0 - 684 MB	NPU	Swin-Small.tflite
Swin-Small	float	Samsung Galaxy S24	Snapdragon® 8 Gen 3 Mobile	TFLITE	11.735 ms	0 - 545 MB	NPU	Swin-Small.tflite
Swin-Small	float	Samsung Galaxy S24	Snapdragon® 8 Gen 3 Mobile	ONNX	10.045 ms	1 - 979 MB	NPU	Swin-Small.onnx.zip
Swin-Small	float	Samsung Galaxy S25	Snapdragon® 8 Elite For Galaxy Mobile	TFLITE	9.047 ms	0 - 408 MB	NPU	Swin-Small.tflite
Swin-Small	float	Samsung Galaxy S25	Snapdragon® 8 Elite For Galaxy Mobile	ONNX	7.723 ms	1 - 656 MB	NPU	Swin-Small.onnx.zip
Swin-Small	float	Snapdragon 8 Elite Gen 5 QRD	Snapdragon® 8 Elite Gen5 Mobile	TFLITE	7.231 ms	0 - 427 MB	NPU	Swin-Small.tflite
Swin-Small	float	Snapdragon 8 Elite Gen 5 QRD	Snapdragon® 8 Elite Gen5 Mobile	ONNX	6.516 ms	1 - 639 MB	NPU	Swin-Small.onnx.zip
Swin-Small	float	Snapdragon X Elite CRD	Snapdragon® X Elite	ONNX	15.982 ms	100 - 100 MB	NPU	Swin-Small.onnx.zip
Swin-Small	w8a16	Dragonwing Q-6690 MTP	Qualcomm® Qcm6690	ONNX	385.414 ms	86 - 107 MB	CPU	Swin-Small.onnx.zip
Swin-Small	w8a16	Dragonwing RB3 Gen 2 Vision Kit	Qualcomm® QCS6490	ONNX	737.745 ms	104 - 124 MB	CPU	Swin-Small.onnx.zip
Swin-Small	w8a16	RB5 (Proxy)	Qualcomm® QCS8250 (Proxy)	ONNX	343.399 ms	100 - 109 MB	CPU	Swin-Small.onnx.zip
Swin-Small	w8a16	Samsung Galaxy S25	Snapdragon® 8 Elite For Galaxy Mobile	ONNX	121.925 ms	36 - 297 MB	NPU	Swin-Small.onnx.zip
Swin-Small	w8a16	Snapdragon 7 Gen 4 QRD	Snapdragon® 7 Gen 4 Mobile	ONNX	357.335 ms	88 - 108 MB	CPU	Swin-Small.onnx.zip
Swin-Small	w8a16	Snapdragon 8 Elite Gen 5 QRD	Snapdragon® 8 Elite Gen5 Mobile	ONNX	123.35 ms	46 - 311 MB	NPU	Swin-Small.onnx.zip
Swin-Small	w8a16	Snapdragon X Elite CRD	Snapdragon® X Elite	ONNX	124.694 ms	65 - 65 MB	NPU	Swin-Small.onnx.zip

Installation

Install the package via pip:

pip install qai-hub-models

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

Sign-in to Qualcomm® AI Hub Workbench 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.

qai-hub configure --api_token API_TOKEN

Navigate to 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.

python -m qai_hub_models.models.swin_small.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.swin_small.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.

python -m qai_hub_models.models.swin_small.export

How does this work?

This export script leverages Qualcomm® AI Hub 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.

import torch

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

# Load the model
torch_model = Model.from_pretrained()

# Device
device = hub.Device("Samsung Galaxy S25")

# 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.

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.

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 Workbench. Sign up for access.

Run demo on a cloud-hosted device

You can also run the demo on-device.

python -m qai_hub_models.models.swin_small.demo --eval-mode 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.swin_small.demo -- --eval-mode on-device

Deploying compiled model to Android

The models can be deployed using multiple runtimes:

• TensorFlow Lite (.tflite export): This tutorial provides a guide to deploy the .tflite model in an Android application.

• QNN (.so export ): This sample app provides instructions on how to use the .so shared library in an Android application.

View on Qualcomm® AI Hub

Get more details on Swin-Small's performance across various devices here. Explore all available models on Qualcomm® AI Hub

License

• The license for the original implementation of Swin-Small can be found here.

References

• Swin Transformer: Hierarchical Vision Transformer using Shifted Windows
• Source Model Implementation

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.