EfficientViT-l2-seg: Optimized for Mobile Deployment

Semantic segmentation in higher resuolution

EfficientViT is a machine learning model that can segment images from the Cityscape dataset. It has lightweight and hardware-efficient operations and thus delivers significant speedup on diverse hardware platforms

This model is an implementation of EfficientViT-l2-seg found here.

This repository provides scripts to run EfficientViT-l2-seg on Qualcomm® devices. More details on model performance across various devices, can be found here.

Model Details

• Model Type: Semantic segmentation
• Model Stats:
  • Model checkpoint: l2.pt
  • Input resolution: 2048x1024
  • Number of parameters: 53M
  • Model size: 200 MB
  • Number of output classes: 19

Model	Device	Chipset	Target Runtime	Inference Time (ms)	Peak Memory Range (MB)	Precision	Primary Compute Unit	Target Model
EfficientViT-l2-seg	Samsung Galaxy S23	Snapdragon® 8 Gen 2	QNN	24783.267 ms	24 - 27 MB	FP16	NPU	EfficientViT-l2-seg.so
EfficientViT-l2-seg	Samsung Galaxy S23	Snapdragon® 8 Gen 2	ONNX	2112.077 ms	5 - 287 MB	FP16	NPU	EfficientViT-l2-seg.onnx
EfficientViT-l2-seg	Samsung Galaxy S24	Snapdragon® 8 Gen 3	QNN	18718.372 ms	26 - 44 MB	FP16	NPU	EfficientViT-l2-seg.so
EfficientViT-l2-seg	Samsung Galaxy S24	Snapdragon® 8 Gen 3	ONNX	2233.69 ms	137 - 885 MB	FP16	NPU	EfficientViT-l2-seg.onnx
EfficientViT-l2-seg	Snapdragon 8 Elite QRD	Snapdragon® 8 Elite	QNN	23012.589 ms	26 - 903 MB	FP16	NPU	Use Export Script
EfficientViT-l2-seg	Snapdragon 8 Elite QRD	Snapdragon® 8 Elite	ONNX	1724.937 ms	96 - 896 MB	FP16	NPU	EfficientViT-l2-seg.onnx
EfficientViT-l2-seg	QCS8275 (Proxy)	QCS8275 Proxy	QNN	51934.305 ms	24 - 31 MB	FP16	NPU	Use Export Script
EfficientViT-l2-seg	QCS8550 (Proxy)	QCS8550 Proxy	QNN	24762.126 ms	24 - 26 MB	FP16	NPU	Use Export Script
EfficientViT-l2-seg	QCS9075 (Proxy)	QCS9075 Proxy	QNN	31821.67 ms	24 - 34 MB	FP16	NPU	Use Export Script
EfficientViT-l2-seg	QCS8450 (Proxy)	QCS8450 Proxy	QNN	25347.273 ms	28 - 1006 MB	FP16	NPU	Use Export Script
EfficientViT-l2-seg	Snapdragon X Elite CRD	Snapdragon® X Elite	QNN	25974.076 ms	24 - 24 MB	FP16	NPU	Use Export Script
EfficientViT-l2-seg	Snapdragon X Elite CRD	Snapdragon® X Elite	ONNX	3097.679 ms	144 - 144 MB	FP16	NPU	EfficientViT-l2-seg.onnx

Installation

Install the package via pip:

pip install "qai-hub-models[efficientvit-l2-seg]"

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

Sign-in to Qualcomm® AI Hub 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.efficientvit_l2_seg.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.efficientvit_l2_seg.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.efficientvit_l2_seg.export

Profiling Results
------------------------------------------------------------
EfficientViT-l2-seg
Device                          : Samsung Galaxy S23 (13)
Runtime                         : QNN                    
Estimated inference time (ms)   : 24783.3                
Estimated peak memory usage (MB): [24, 27]               
Total # Ops                     : 775                    
Compute Unit(s)                 : NPU (775 ops)          

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.efficientvit_l2_seg 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.

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. 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.efficientvit_l2_seg.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.efficientvit_l2_seg.demo -- --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 EfficientViT-l2-seg's performance across various devices here. Explore all available models on Qualcomm® AI Hub

License

• The license for the original implementation of EfficientViT-l2-seg can be found here.
• The license for the compiled assets for on-device deployment can be found here

References

• EfficientViT: Multi-Scale Linear Attention for High-Resolution Dense Prediction
• 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.