Hugging Face's logo

qualcomm
/
MobileSam

Image Segmentation
PyTorch
foundation
android
Model card Files
xet
Community
MobileSam / README.md
qaihm-bot's picture
qaihm-bot
Upload README.md with huggingface_hub
20d1843 verified
preview code
|
raw
history blame
17.7 kB
 ---
library_name: pytorch
license: other
tags:
- foundation
- android
pipeline_tag: image-segmentation
---
![](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/mobilesam/web-assets/model_demo.png)
# 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](https://github.com/facebookresearch/segment-anything).
This repository provides scripts to run MobileSam on Qualcomm® devices.
More details on model performance across various devices, can be found
[here](https://aihub.qualcomm.com/models/mobilesam).
### Model Details
- **Model Type:** Model_use_case.semantic_segmentation
- **Model Stats:**
- Model checkpoint: vit_t
- Input resolution: 720p (720x1280)
- Number of parameters (MobileSamDecoder): 3.876M
- Model size (MobileSamDecoder): 19.6 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 | 769.353 ms | 33 - 165 MB | NPU | [MobileSam.tflite](https://huggingface.co/qualcomm/MobileSam/blob/main/MobileSam.tflite) |
| SAMEncoder | float | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | QNN | 501.318 ms | 3 - 13 MB | NPU | Use Export Script |
| SAMEncoder | float | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | TFLITE | 588.112 ms | 33 - 175 MB | NPU | [MobileSam.tflite](https://huggingface.co/qualcomm/MobileSam/blob/main/MobileSam.tflite) |
| SAMEncoder | float | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | QNN | 477.804 ms | 12 - 587 MB | NPU | Use Export Script |
| SAMEncoder | float | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | TFLITE | 441.504 ms | 33 - 60 MB | NPU | [MobileSam.tflite](https://huggingface.co/qualcomm/MobileSam/blob/main/MobileSam.tflite) |
| SAMEncoder | float | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | QNN | 268.611 ms | 12 - 14 MB | NPU | Use Export Script |
| SAMEncoder | float | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | TFLITE | 420.17 ms | 33 - 166 MB | NPU | [MobileSam.tflite](https://huggingface.co/qualcomm/MobileSam/blob/main/MobileSam.tflite) |
| SAMEncoder | float | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | QNN | 274.61 ms | 2 - 16 MB | NPU | Use Export Script |
| SAMEncoder | float | SA7255P ADP | Qualcomm® SA7255P | TFLITE | 769.353 ms | 33 - 165 MB | NPU | [MobileSam.tflite](https://huggingface.co/qualcomm/MobileSam/blob/main/MobileSam.tflite) |
| SAMEncoder | float | SA7255P ADP | Qualcomm® SA7255P | QNN | 501.318 ms | 3 - 13 MB | NPU | Use Export Script |
| SAMEncoder | float | SA8255 (Proxy) | Qualcomm® SA8255P (Proxy) | TFLITE | 429.093 ms | 33 - 68 MB | NPU | [MobileSam.tflite](https://huggingface.co/qualcomm/MobileSam/blob/main/MobileSam.tflite) |
| SAMEncoder | float | SA8255 (Proxy) | Qualcomm® SA8255P (Proxy) | QNN | 269.116 ms | 3 - 5 MB | NPU | Use Export Script |
| SAMEncoder | float | SA8295P ADP | Qualcomm® SA8295P | TFLITE | 586.886 ms | 28 - 163 MB | NPU | [MobileSam.tflite](https://huggingface.co/qualcomm/MobileSam/blob/main/MobileSam.tflite) |
| SAMEncoder | float | SA8295P ADP | Qualcomm® SA8295P | QNN | 424.964 ms | 0 - 18 MB | NPU | Use Export Script |
| SAMEncoder | float | SA8650 (Proxy) | Qualcomm® SA8650P (Proxy) | TFLITE | 421.365 ms | 33 - 60 MB | NPU | [MobileSam.tflite](https://huggingface.co/qualcomm/MobileSam/blob/main/MobileSam.tflite) |
| SAMEncoder | float | SA8650 (Proxy) | Qualcomm® SA8650P (Proxy) | QNN | 267.29 ms | 12 - 14 MB | NPU | Use Export Script |
| SAMEncoder | float | SA8775P ADP | Qualcomm® SA8775P | TFLITE | 420.17 ms | 33 - 166 MB | NPU | [MobileSam.tflite](https://huggingface.co/qualcomm/MobileSam/blob/main/MobileSam.tflite) |
| SAMEncoder | float | SA8775P ADP | Qualcomm® SA8775P | QNN | 274.61 ms | 2 - 16 MB | NPU | Use Export Script |
| SAMEncoder | float | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 Mobile | TFLITE | 419.834 ms | 33 - 59 MB | NPU | [MobileSam.tflite](https://huggingface.co/qualcomm/MobileSam/blob/main/MobileSam.tflite) |
| SAMEncoder | float | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 Mobile | QNN | 270.005 ms | 12 - 87 MB | NPU | Use Export Script |
| SAMEncoder | float | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 Mobile | ONNX | 398.739 ms | 52 - 139 MB | NPU | [MobileSam.onnx](https://huggingface.co/qualcomm/MobileSam/blob/main/MobileSam.onnx) |
| SAMEncoder | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | TFLITE | 333.791 ms | 33 - 154 MB | NPU | [MobileSam.tflite](https://huggingface.co/qualcomm/MobileSam/blob/main/MobileSam.tflite) |
| SAMEncoder | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | QNN | 202.059 ms | 12 - 620 MB | NPU | Use Export Script |
| SAMEncoder | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | ONNX | 298.296 ms | 96 - 213 MB | NPU | [MobileSam.onnx](https://huggingface.co/qualcomm/MobileSam/blob/main/MobileSam.onnx) |
| SAMEncoder | float | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite Mobile | TFLITE | 277.237 ms | 32 - 165 MB | NPU | [MobileSam.tflite](https://huggingface.co/qualcomm/MobileSam/blob/main/MobileSam.tflite) |
| SAMEncoder | float | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite Mobile | QNN | 175.191 ms | 12 - 599 MB | NPU | Use Export Script |
| SAMEncoder | float | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite Mobile | ONNX | 223.436 ms | 92 - 220 MB | NPU | [MobileSam.onnx](https://huggingface.co/qualcomm/MobileSam/blob/main/MobileSam.onnx) |
| SAMEncoder | float | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN | 275.768 ms | 12 - 12 MB | NPU | Use Export Script |
| SAMEncoder | float | Snapdragon X Elite CRD | Snapdragon® X Elite | ONNX | 449.783 ms | 130 - 130 MB | NPU | [MobileSam.onnx](https://huggingface.co/qualcomm/MobileSam/blob/main/MobileSam.onnx) |
| SAMDecoder | float | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | TFLITE | 16.776 ms | 0 - 40 MB | NPU | [MobileSam.tflite](https://huggingface.co/qualcomm/MobileSam/blob/main/MobileSam.tflite) |
| SAMDecoder | float | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | QNN | 13.918 ms | 1 - 12 MB | NPU | Use Export Script |
| SAMDecoder | float | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | TFLITE | 8.905 ms | 0 - 42 MB | NPU | [MobileSam.tflite](https://huggingface.co/qualcomm/MobileSam/blob/main/MobileSam.tflite) |
| SAMDecoder | float | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | QNN | 8.838 ms | 4 - 49 MB | NPU | Use Export Script |
| SAMDecoder | float | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | TFLITE | 7.355 ms | 0 - 29 MB | NPU | [MobileSam.tflite](https://huggingface.co/qualcomm/MobileSam/blob/main/MobileSam.tflite) |
| SAMDecoder | float | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | QNN | 6.17 ms | 4 - 6 MB | NPU | Use Export Script |
| SAMDecoder | float | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | TFLITE | 8.588 ms | 0 - 41 MB | NPU | [MobileSam.tflite](https://huggingface.co/qualcomm/MobileSam/blob/main/MobileSam.tflite) |
| SAMDecoder | float | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | QNN | 7.343 ms | 2 - 16 MB | NPU | Use Export Script |
| SAMDecoder | float | SA7255P ADP | Qualcomm® SA7255P | TFLITE | 16.776 ms | 0 - 40 MB | NPU | [MobileSam.tflite](https://huggingface.co/qualcomm/MobileSam/blob/main/MobileSam.tflite) |
| SAMDecoder | float | SA7255P ADP | Qualcomm® SA7255P | QNN | 13.918 ms | 1 - 12 MB | NPU | Use Export Script |
| SAMDecoder | float | SA8255 (Proxy) | Qualcomm® SA8255P (Proxy) | TFLITE | 7.349 ms | 0 - 31 MB | NPU | [MobileSam.tflite](https://huggingface.co/qualcomm/MobileSam/blob/main/MobileSam.tflite) |
| SAMDecoder | float | SA8255 (Proxy) | Qualcomm® SA8255P (Proxy) | QNN | 6.195 ms | 5 - 7 MB | NPU | Use Export Script |
| SAMDecoder | float | SA8295P ADP | Qualcomm® SA8295P | TFLITE | 9.829 ms | 0 - 36 MB | NPU | [MobileSam.tflite](https://huggingface.co/qualcomm/MobileSam/blob/main/MobileSam.tflite) |
| SAMDecoder | float | SA8295P ADP | Qualcomm® SA8295P | QNN | 7.385 ms | 0 - 18 MB | NPU | Use Export Script |
| SAMDecoder | float | SA8650 (Proxy) | Qualcomm® SA8650P (Proxy) | TFLITE | 7.375 ms | 0 - 29 MB | NPU | [MobileSam.tflite](https://huggingface.co/qualcomm/MobileSam/blob/main/MobileSam.tflite) |
| SAMDecoder | float | SA8650 (Proxy) | Qualcomm® SA8650P (Proxy) | QNN | 6.194 ms | 4 - 6 MB | NPU | Use Export Script |
| SAMDecoder | float | SA8775P ADP | Qualcomm® SA8775P | TFLITE | 8.588 ms | 0 - 41 MB | NPU | [MobileSam.tflite](https://huggingface.co/qualcomm/MobileSam/blob/main/MobileSam.tflite) |
| SAMDecoder | float | SA8775P ADP | Qualcomm® SA8775P | QNN | 7.343 ms | 2 - 16 MB | NPU | Use Export Script |
| SAMDecoder | float | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 Mobile | TFLITE | 7.341 ms | 0 - 32 MB | NPU | [MobileSam.tflite](https://huggingface.co/qualcomm/MobileSam/blob/main/MobileSam.tflite) |
| SAMDecoder | float | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 Mobile | QNN | 6.126 ms | 4 - 19 MB | NPU | Use Export Script |
| SAMDecoder | float | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 Mobile | ONNX | 8.874 ms | 1 - 65 MB | NPU | [MobileSam.onnx](https://huggingface.co/qualcomm/MobileSam/blob/main/MobileSam.onnx) |
| SAMDecoder | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | TFLITE | 5.163 ms | 0 - 49 MB | NPU | [MobileSam.tflite](https://huggingface.co/qualcomm/MobileSam/blob/main/MobileSam.tflite) |
| SAMDecoder | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | QNN | 4.138 ms | 4 - 52 MB | NPU | Use Export Script |
| SAMDecoder | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | ONNX | 5.9 ms | 4 - 70 MB | NPU | [MobileSam.onnx](https://huggingface.co/qualcomm/MobileSam/blob/main/MobileSam.onnx) |
| SAMDecoder | float | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite Mobile | TFLITE | 5.025 ms | 0 - 44 MB | NPU | [MobileSam.tflite](https://huggingface.co/qualcomm/MobileSam/blob/main/MobileSam.tflite) |
| SAMDecoder | float | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite Mobile | QNN | 3.793 ms | 4 - 44 MB | NPU | Use Export Script |
| SAMDecoder | float | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite Mobile | ONNX | 5.397 ms | 3 - 61 MB | NPU | [MobileSam.onnx](https://huggingface.co/qualcomm/MobileSam/blob/main/MobileSam.onnx) |
| SAMDecoder | float | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN | 6.692 ms | 4 - 4 MB | NPU | Use Export Script |
| SAMDecoder | float | Snapdragon X Elite CRD | Snapdragon® X Elite | ONNX | 10.02 ms | 12 - 12 MB | NPU | [MobileSam.onnx](https://huggingface.co/qualcomm/MobileSam/blob/main/MobileSam.onnx) |
## Installation
Install the package via pip:
```bash
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](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.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.
```bash
python -m qai_hub_models.models.mobilesam.export
```
```
Profiling Results
------------------------------------------------------------
SAMEncoder
Device : cs_8275 (ANDROID 14)
Runtime : TFLITE
Estimated inference time (ms) : 769.4
Estimated peak memory usage (MB): [33, 165]
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) : 16.8
Estimated peak memory usage (MB): [0, 40]
Total # Ops : 845
Compute Unit(s) : npu (845 ops) gpu (0 ops) cpu (0 ops)
```
## How does this work?
This [export script](https://aihub.qualcomm.com/models/mobilesam/qai_hub_models/models/MobileSam/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.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.
```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.mobilesam.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.mobilesam.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 MobileSam's performance across various devices [here](https://aihub.qualcomm.com/models/mobilesam).
Explore all available models on [Qualcomm® AI Hub](https://aihub.qualcomm.com/)
## License
* The license for the original implementation of MobileSam can be found
[here](https://github.com/facebookresearch/segment-anything/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
* [Segment Anything](https://arxiv.org/abs/2306.14289)
* [Source Model Implementation](https://github.com/facebookresearch/segment-anything)
## 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).