Segment-Anything-Model: Optimized for Mobile Deployment

High-quality segmentation mask generation around any object in an image with simple input prompt

Transformer based encoder-decoder where prompts specify what to segment in an image thereby allowing segmentation without the need for additional training. The image encoder generates embeddings and the lightweight decoder operates on the embeddings for point and mask based image segmentation.

This model is an implementation of Segment-Anything-Model found here.

This repository provides scripts to run Segment-Anything-Model 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: vit_l
- Input resolution: 720p (720x1280)
- Number of parameters (SAMDecoder): 5.11M
- Model size (SAMDecoder): 19.6 MB

Model	Device	Chipset	Target Runtime	Inference Time (ms)	Peak Memory Range (MB)	Precision	Primary Compute Unit	Target Model
SAMDecoder	Samsung Galaxy S23	Snapdragon® 8 Gen 2	TFLITE	29.311 ms	4 - 40 MB	FP16	NPU	Segment-Anything-Model.tflite
SAMDecoder	Samsung Galaxy S24	Snapdragon® 8 Gen 3	TFLITE	20.843 ms	4 - 119 MB	FP16	NPU	Segment-Anything-Model.tflite
SAMDecoder	Snapdragon 8 Elite QRD	Snapdragon® 8 Elite	TFLITE	20.649 ms	3 - 128 MB	FP16	NPU	Segment-Anything-Model.tflite
SAMDecoder	QCS8550 (Proxy)	QCS8550 Proxy	TFLITE	29.453 ms	4 - 41 MB	FP16	NPU	Segment-Anything-Model.tflite
SAMDecoder	SA8255 (Proxy)	SA8255P Proxy	TFLITE	29.485 ms	4 - 36 MB	FP16	NPU	Segment-Anything-Model.tflite
SAMDecoder	SA8295P ADP	SA8295P	TFLITE	36.583 ms	4 - 114 MB	FP16	NPU	Segment-Anything-Model.tflite
SAMDecoder	SA8650 (Proxy)	SA8650P Proxy	TFLITE	29.064 ms	4 - 37 MB	FP16	NPU	Segment-Anything-Model.tflite
SAMDecoder	QCS8450 (Proxy)	QCS8450 Proxy	TFLITE	33.135 ms	4 - 126 MB	FP16	NPU	Segment-Anything-Model.tflite
SAMEncoder	Samsung Galaxy S23	Snapdragon® 8 Gen 2	TFLITE	10369.27 ms	123 - 126 MB	FP32	CPU	Segment-Anything-Model.tflite
SAMEncoder	Samsung Galaxy S24	Snapdragon® 8 Gen 3	TFLITE	9136.859 ms	91 - 1603 MB	FP32	CPU	Segment-Anything-Model.tflite
SAMEncoder	Snapdragon 8 Elite QRD	Snapdragon® 8 Elite	TFLITE	6694.41 ms	110 - 1589 MB	FP32	CPU	Segment-Anything-Model.tflite
SAMEncoder	QCS8550 (Proxy)	QCS8550 Proxy	TFLITE	10714.354 ms	122 - 381 MB	FP32	CPU	Segment-Anything-Model.tflite
SAMEncoder	SA7255P ADP	SA7255P	TFLITE	17192.024 ms	125 - 1597 MB	FP32	CPU	Segment-Anything-Model.tflite
SAMEncoder	SA8255 (Proxy)	SA8255P Proxy	TFLITE	11081.197 ms	124 - 127 MB	FP32	CPU	Segment-Anything-Model.tflite
SAMEncoder	SA8295P ADP	SA8295P	TFLITE	10359.771 ms	124 - 1654 MB	FP32	CPU	Segment-Anything-Model.tflite
SAMEncoder	SA8650 (Proxy)	SA8650P Proxy	TFLITE	10872.614 ms	118 - 125 MB	FP32	CPU	Segment-Anything-Model.tflite
SAMEncoder	SA8775P ADP	SA8775P	TFLITE	13361.634 ms	124 - 1596 MB	FP32	CPU	Segment-Anything-Model.tflite
SAMEncoder	QCS8450 (Proxy)	QCS8450 Proxy	TFLITE	16830.94 ms	82 - 1648 MB	FP32	CPU	Segment-Anything-Model.tflite

Installation

This model can be installed as a Python package via pip.

pip install "qai-hub-models[sam]"

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.sam.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.sam.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.sam.export

Profiling Results
------------------------------------------------------------
SAMDecoder
Device                          : Samsung Galaxy S23 (13)
Runtime                         : TFLITE                 
Estimated inference time (ms)   : 29.3                   
Estimated peak memory usage (MB): [4, 40]                
Total # Ops                     : 337                    
Compute Unit(s)                 : NPU (337 ops)          

------------------------------------------------------------
SAMEncoder
Device                          : Samsung Galaxy S23 (13)   
Runtime                         : TFLITE                    
Estimated inference time (ms)   : 10369.3                   
Estimated peak memory usage (MB): [123, 126]                
Total # Ops                     : 818                       
Compute Unit(s)                 : GPU (36 ops) CPU (782 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.sam import Model

# Load the model
model = Model.from_pretrained()
get_sam_decoder()_model = model.get_sam_decoder()
get_sam_encoder()_model = model.get_sam_encoder()

# Device
device = hub.Device("Samsung Galaxy S23")

# Trace model
get_sam_decoder()_input_shape = get_sam_decoder()_model.get_input_spec()
get_sam_decoder()_sample_inputs = get_sam_decoder()_model.sample_inputs()

traced_get_sam_decoder()_model = torch.jit.trace(get_sam_decoder()_model, [torch.tensor(data[0]) for _, data in get_sam_decoder()_sample_inputs.items()])

# Compile model on a specific device
get_sam_decoder()_compile_job = hub.submit_compile_job(
    model=traced_get_sam_decoder()_model ,
    device=device,
    input_specs=get_sam_decoder()_model.get_input_spec(),
)

# Get target model to run on-device
get_sam_decoder()_target_model = get_sam_decoder()_compile_job.get_target_model()
# Trace model
get_sam_encoder()_input_shape = get_sam_encoder()_model.get_input_spec()
get_sam_encoder()_sample_inputs = get_sam_encoder()_model.sample_inputs()

traced_get_sam_encoder()_model = torch.jit.trace(get_sam_encoder()_model, [torch.tensor(data[0]) for _, data in get_sam_encoder()_sample_inputs.items()])

# Compile model on a specific device
get_sam_encoder()_compile_job = hub.submit_compile_job(
    model=traced_get_sam_encoder()_model ,
    device=device,
    input_specs=get_sam_encoder()_model.get_input_spec(),
)

# Get target model to run on-device
get_sam_encoder()_target_model = get_sam_encoder()_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.

get_sam_decoder()_profile_job = hub.submit_profile_job(
    model=get_sam_decoder()_target_model,
    device=device,
)
get_sam_encoder()_profile_job = hub.submit_profile_job(
    model=get_sam_encoder()_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.

get_sam_decoder()_input_data = get_sam_decoder()_model.sample_inputs()
get_sam_decoder()_inference_job = hub.submit_inference_job(
    model=get_sam_decoder()_target_model,
    device=device,
    inputs=get_sam_decoder()_input_data,
)
get_sam_decoder()_inference_job.download_output_data()
get_sam_encoder()_input_data = get_sam_encoder()_model.sample_inputs()
get_sam_encoder()_inference_job = hub.submit_inference_job(
    model=get_sam_encoder()_target_model,
    device=device,
    inputs=get_sam_encoder()_input_data,
)
get_sam_encoder()_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.sam.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.sam.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 Segment-Anything-Model's performance across various devices here. Explore all available models on Qualcomm® AI Hub

License

The license for the original implementation of Segment-Anything-Model can be found here.
The license for the compiled assets for on-device deployment can be found here

References

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.