Hugging Face's logo

qualcomm
/
MobileSam

Image Segmentation
PyTorch
LiteRT
ONNX
foundation
android
Model card Files Community

MobileSam: Optimized for Mobile Deployment

Faster Segment Anything: Towards lightweight SAM for mobile applications

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 MobileSam found here.

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

Model Details

  • Model Type: Model_use_case.semantic_segmentation
  • Model Stats:
    • Model checkpoint: vit_t
    • Input resolution: 720p (720x1280)
    • Number of parameters (SAMEncoder): 6.95M
    • Model size (SAMEncoder) (float): 26.6 MB
    • Number of parameters (SAMDecoder): 6.16M
    • Model size (SAMDecoder) (float): 23.7 MB
Model Precision Device Chipset Target Runtime Inference Time (ms) Peak Memory Range (MB) Primary Compute Unit Target Model
SAMEncoder float QCS8275 (Proxy) Qualcomm® QCS8275 (Proxy) TFLITE 763.405 ms 33 - 381 MB NPU MobileSam.tflite
SAMEncoder float QCS8275 (Proxy) Qualcomm® QCS8275 (Proxy) QNN_DLC 510.611 ms 5 - 1161 MB NPU MobileSam.dlc
SAMEncoder float QCS8450 (Proxy) Qualcomm® QCS8450 (Proxy) TFLITE 626.456 ms 33 - 731 MB NPU MobileSam.tflite
SAMEncoder float QCS8450 (Proxy) Qualcomm® QCS8450 (Proxy) QNN_DLC 478.836 ms 12 - 618 MB NPU MobileSam.dlc
SAMEncoder float QCS8550 (Proxy) Qualcomm® QCS8550 (Proxy) TFLITE 434.778 ms 33 - 58 MB NPU MobileSam.tflite
SAMEncoder float QCS8550 (Proxy) Qualcomm® QCS8550 (Proxy) QNN_DLC 272.46 ms 12 - 83 MB NPU MobileSam.dlc
SAMEncoder float QCS9075 (Proxy) Qualcomm® QCS9075 (Proxy) TFLITE 423.426 ms 33 - 379 MB NPU MobileSam.tflite
SAMEncoder float QCS9075 (Proxy) Qualcomm® QCS9075 (Proxy) QNN_DLC 277.611 ms 1 - 1161 MB NPU MobileSam.dlc
SAMEncoder float SA7255P ADP Qualcomm® SA7255P TFLITE 763.405 ms 33 - 381 MB NPU MobileSam.tflite
SAMEncoder float SA7255P ADP Qualcomm® SA7255P QNN_DLC 510.611 ms 5 - 1161 MB NPU MobileSam.dlc
SAMEncoder float SA8255 (Proxy) Qualcomm® SA8255P (Proxy) TFLITE 441.613 ms 33 - 76 MB NPU MobileSam.tflite
SAMEncoder float SA8255 (Proxy) Qualcomm® SA8255P (Proxy) QNN_DLC 273.174 ms 12 - 88 MB NPU MobileSam.dlc
SAMEncoder float SA8295P ADP Qualcomm® SA8295P TFLITE 586.065 ms 33 - 382 MB NPU MobileSam.tflite
SAMEncoder float SA8295P ADP Qualcomm® SA8295P QNN_DLC 436.442 ms 5 - 651 MB NPU MobileSam.dlc
SAMEncoder float SA8650 (Proxy) Qualcomm® SA8650P (Proxy) TFLITE 437.511 ms 33 - 61 MB NPU MobileSam.tflite
SAMEncoder float SA8650 (Proxy) Qualcomm® SA8650P (Proxy) QNN_DLC 271.999 ms 12 - 89 MB NPU MobileSam.dlc
SAMEncoder float SA8775P ADP Qualcomm® SA8775P TFLITE 423.426 ms 33 - 379 MB NPU MobileSam.tflite
SAMEncoder float SA8775P ADP Qualcomm® SA8775P QNN_DLC 277.611 ms 1 - 1161 MB NPU MobileSam.dlc
SAMEncoder float Samsung Galaxy S23 Snapdragon® 8 Gen 2 Mobile TFLITE 443.612 ms 25 - 58 MB NPU MobileSam.tflite
SAMEncoder float Samsung Galaxy S23 Snapdragon® 8 Gen 2 Mobile QNN_DLC 272.764 ms 12 - 74 MB NPU MobileSam.dlc
SAMEncoder float Samsung Galaxy S23 Snapdragon® 8 Gen 2 Mobile ONNX 417.653 ms 95 - 152 MB NPU MobileSam.onnx
SAMEncoder float Samsung Galaxy S24 Snapdragon® 8 Gen 3 Mobile TFLITE 339.38 ms 33 - 712 MB NPU MobileSam.tflite
SAMEncoder float Samsung Galaxy S24 Snapdragon® 8 Gen 3 Mobile QNN_DLC 206.041 ms 12 - 2423 MB NPU MobileSam.dlc
SAMEncoder float Samsung Galaxy S24 Snapdragon® 8 Gen 3 Mobile ONNX 303.821 ms 97 - 827 MB NPU MobileSam.onnx
SAMEncoder float Snapdragon 8 Elite QRD Snapdragon® 8 Elite Mobile TFLITE 236.15 ms 31 - 378 MB NPU MobileSam.tflite
SAMEncoder float Snapdragon 8 Elite QRD Snapdragon® 8 Elite Mobile QNN_DLC 173.015 ms 12 - 1181 MB NPU MobileSam.dlc
SAMEncoder float Snapdragon 8 Elite QRD Snapdragon® 8 Elite Mobile ONNX 255.315 ms 99 - 502 MB NPU MobileSam.onnx
SAMEncoder float Snapdragon X Elite CRD Snapdragon® X Elite QNN_DLC 298.206 ms 85 - 85 MB NPU MobileSam.dlc
SAMEncoder float Snapdragon X Elite CRD Snapdragon® X Elite ONNX 436.531 ms 131 - 131 MB NPU MobileSam.onnx
SAMDecoder float QCS8275 (Proxy) Qualcomm® QCS8275 (Proxy) TFLITE 16.509 ms 0 - 55 MB NPU MobileSam.tflite
SAMDecoder float QCS8275 (Proxy) Qualcomm® QCS8275 (Proxy) QNN_DLC 13.893 ms 4 - 47 MB NPU MobileSam.dlc
SAMDecoder float QCS8450 (Proxy) Qualcomm® QCS8450 (Proxy) TFLITE 8.268 ms 0 - 55 MB NPU MobileSam.tflite
SAMDecoder float QCS8450 (Proxy) Qualcomm® QCS8450 (Proxy) QNN_DLC 8.385 ms 4 - 54 MB NPU MobileSam.dlc
SAMDecoder float QCS8550 (Proxy) Qualcomm® QCS8550 (Proxy) TFLITE 7.236 ms 0 - 30 MB NPU MobileSam.tflite
SAMDecoder float QCS8550 (Proxy) Qualcomm® QCS8550 (Proxy) QNN_DLC 6.11 ms 4 - 21 MB NPU MobileSam.dlc
SAMDecoder float QCS9075 (Proxy) Qualcomm® QCS9075 (Proxy) TFLITE 8.435 ms 0 - 54 MB NPU MobileSam.tflite
SAMDecoder float QCS9075 (Proxy) Qualcomm® QCS9075 (Proxy) QNN_DLC 7.283 ms 1 - 44 MB NPU MobileSam.dlc
SAMDecoder float SA7255P ADP Qualcomm® SA7255P TFLITE 16.509 ms 0 - 55 MB NPU MobileSam.tflite
SAMDecoder float SA7255P ADP Qualcomm® SA7255P QNN_DLC 13.893 ms 4 - 47 MB NPU MobileSam.dlc
SAMDecoder float SA8255 (Proxy) Qualcomm® SA8255P (Proxy) TFLITE 7.262 ms 0 - 30 MB NPU MobileSam.tflite
SAMDecoder float SA8255 (Proxy) Qualcomm® SA8255P (Proxy) QNN_DLC 6.084 ms 4 - 21 MB NPU MobileSam.dlc
SAMDecoder float SA8295P ADP Qualcomm® SA8295P TFLITE 9.689 ms 0 - 50 MB NPU MobileSam.tflite
SAMDecoder float SA8295P ADP Qualcomm® SA8295P QNN_DLC 7.485 ms 0 - 59 MB NPU MobileSam.dlc
SAMDecoder float SA8650 (Proxy) Qualcomm® SA8650P (Proxy) TFLITE 7.3 ms 0 - 29 MB NPU MobileSam.tflite
SAMDecoder float SA8650 (Proxy) Qualcomm® SA8650P (Proxy) QNN_DLC 6.095 ms 4 - 20 MB NPU MobileSam.dlc
SAMDecoder float SA8775P ADP Qualcomm® SA8775P TFLITE 8.435 ms 0 - 54 MB NPU MobileSam.tflite
SAMDecoder float SA8775P ADP Qualcomm® SA8775P QNN_DLC 7.283 ms 1 - 44 MB NPU MobileSam.dlc
SAMDecoder float Samsung Galaxy S23 Snapdragon® 8 Gen 2 Mobile TFLITE 7.277 ms 0 - 31 MB NPU MobileSam.tflite
SAMDecoder float Samsung Galaxy S23 Snapdragon® 8 Gen 2 Mobile QNN_DLC 6.101 ms 4 - 24 MB NPU MobileSam.dlc
SAMDecoder float Samsung Galaxy S23 Snapdragon® 8 Gen 2 Mobile ONNX 9.654 ms 1 - 66 MB NPU MobileSam.onnx
SAMDecoder float Samsung Galaxy S24 Snapdragon® 8 Gen 3 Mobile TFLITE 5.104 ms 0 - 61 MB NPU MobileSam.tflite
SAMDecoder float Samsung Galaxy S24 Snapdragon® 8 Gen 3 Mobile QNN_DLC 4.201 ms 0 - 55 MB NPU MobileSam.dlc
SAMDecoder float Samsung Galaxy S24 Snapdragon® 8 Gen 3 Mobile ONNX 6.1 ms 4 - 75 MB NPU MobileSam.onnx
SAMDecoder float Snapdragon 8 Elite QRD Snapdragon® 8 Elite Mobile TFLITE 5.075 ms 0 - 57 MB NPU MobileSam.tflite
SAMDecoder float Snapdragon 8 Elite QRD Snapdragon® 8 Elite Mobile QNN_DLC 3.695 ms 4 - 55 MB NPU MobileSam.dlc
SAMDecoder float Snapdragon 8 Elite QRD Snapdragon® 8 Elite Mobile ONNX 5.362 ms 4 - 65 MB NPU MobileSam.onnx
SAMDecoder float Snapdragon X Elite CRD Snapdragon® X Elite QNN_DLC 7.965 ms 4 - 4 MB NPU MobileSam.dlc
SAMDecoder float Snapdragon X Elite CRD Snapdragon® X Elite ONNX 10.148 ms 11 - 11 MB NPU MobileSam.onnx

Installation

Install the package via pip:

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

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.mobilesam.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.mobilesam.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.mobilesam.export

Profiling Results
------------------------------------------------------------
SAMEncoder
Device                          : cs_8275 (ANDROID 14)                  
Runtime                         : TFLITE                                
Estimated inference time (ms)   : 763.4                                 
Estimated peak memory usage (MB): [33, 381]                             
Total # Ops                     : 592                                   
Compute Unit(s)                 : npu (532 ops) gpu (0 ops) cpu (60 ops)

------------------------------------------------------------
SAMDecoder
Device                          : cs_8275 (ANDROID 14)                 
Runtime                         : TFLITE                               
Estimated inference time (ms)   : 16.5                                 
Estimated peak memory usage (MB): [0, 55]                              
Total # Ops                     : 846                                  
Compute Unit(s)                 : npu (846 ops) gpu (0 ops) cpu (0 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.mobilesam 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.mobilesam.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.mobilesam.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 MobileSam's performance across various devices here. Explore all available models on Qualcomm® AI Hub

License

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

References

Community

Downloads last month
52
Inference Providers NEW
Image Segmentation
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support