--- library_name: pytorch license: agpl-3.0 pipeline_tag: image-segmentation tags: - android --- ![](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/fastsam_x/web-assets/model_demo.png) # FastSam-X: Optimized for Mobile Deployment ## Generate high quality segmentation mask on device The Fast Segment Anything Model (FastSAM) is a novel, real-time CNN-based solution for the Segment Anything task. This task is designed to segment any object within an image based on various possible user interaction prompts. The model performs competitively despite significantly reduced computation, making it a practical choice for a variety of vision tasks. This model is an implementation of FastSam-X found [here](https://github.com/CASIA-IVA-Lab/FastSAM). This repository provides scripts to run FastSam-X on Qualcomm® devices. More details on model performance across various devices, can be found [here](https://aihub.qualcomm.com/models/fastsam_x). ### Model Details - **Model Type:** Semantic segmentation - **Model Stats:** - Model checkpoint: fastsam-x.pt - Inference latency: RealTime - Input resolution: 640x640 - Number of parameters: 72.2M - Model size: 276 MB | Model | Device | Chipset | Target Runtime | Inference Time (ms) | Peak Memory Range (MB) | Precision | Primary Compute Unit | Target Model |---|---|---|---|---|---|---|---|---| | FastSam-X | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 | TFLITE | 46.279 ms | 4 - 19 MB | FP16 | NPU | [FastSam-X.tflite](https://huggingface.co/qualcomm/FastSam-X/blob/main/FastSam-X.tflite) | | FastSam-X | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 | QNN | 45.042 ms | 4 - 21 MB | FP16 | NPU | [FastSam-X.so](https://huggingface.co/qualcomm/FastSam-X/blob/main/FastSam-X.so) | | FastSam-X | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 | ONNX | 49.854 ms | 5 - 318 MB | FP16 | NPU | [FastSam-X.onnx](https://huggingface.co/qualcomm/FastSam-X/blob/main/FastSam-X.onnx) | | FastSam-X | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | TFLITE | 34.909 ms | 1 - 61 MB | FP16 | NPU | [FastSam-X.tflite](https://huggingface.co/qualcomm/FastSam-X/blob/main/FastSam-X.tflite) | | FastSam-X | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | QNN | 34.417 ms | 5 - 64 MB | FP16 | NPU | [FastSam-X.so](https://huggingface.co/qualcomm/FastSam-X/blob/main/FastSam-X.so) | | FastSam-X | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | ONNX | 34.958 ms | 14 - 79 MB | FP16 | NPU | [FastSam-X.onnx](https://huggingface.co/qualcomm/FastSam-X/blob/main/FastSam-X.onnx) | | FastSam-X | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | TFLITE | 27.589 ms | 4 - 64 MB | FP16 | NPU | [FastSam-X.tflite](https://huggingface.co/qualcomm/FastSam-X/blob/main/FastSam-X.tflite) | | FastSam-X | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | QNN | 27.389 ms | 5 - 64 MB | FP16 | NPU | Use Export Script | | FastSam-X | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | ONNX | 31.897 ms | 21 - 76 MB | FP16 | NPU | [FastSam-X.onnx](https://huggingface.co/qualcomm/FastSam-X/blob/main/FastSam-X.onnx) | | FastSam-X | QCS8550 (Proxy) | QCS8550 Proxy | TFLITE | 45.14 ms | 4 - 56 MB | FP16 | NPU | [FastSam-X.tflite](https://huggingface.co/qualcomm/FastSam-X/blob/main/FastSam-X.tflite) | | FastSam-X | QCS8550 (Proxy) | QCS8550 Proxy | QNN | 45.059 ms | 5 - 8 MB | FP16 | NPU | Use Export Script | | FastSam-X | SA7255P ADP | SA7255P | TFLITE | 2097.686 ms | 0 - 60 MB | FP16 | NPU | [FastSam-X.tflite](https://huggingface.co/qualcomm/FastSam-X/blob/main/FastSam-X.tflite) | | FastSam-X | SA7255P ADP | SA7255P | QNN | 2095.613 ms | 5 - 14 MB | FP16 | NPU | Use Export Script | | FastSam-X | SA8255 (Proxy) | SA8255P Proxy | TFLITE | 46.078 ms | 4 - 51 MB | FP16 | NPU | [FastSam-X.tflite](https://huggingface.co/qualcomm/FastSam-X/blob/main/FastSam-X.tflite) | | FastSam-X | SA8255 (Proxy) | SA8255P Proxy | QNN | 44.533 ms | 5 - 7 MB | FP16 | NPU | Use Export Script | | FastSam-X | SA8295P ADP | SA8295P | TFLITE | 93.938 ms | 0 - 56 MB | FP16 | NPU | [FastSam-X.tflite](https://huggingface.co/qualcomm/FastSam-X/blob/main/FastSam-X.tflite) | | FastSam-X | SA8295P ADP | SA8295P | QNN | 82.809 ms | 0 - 14 MB | FP16 | NPU | Use Export Script | | FastSam-X | SA8650 (Proxy) | SA8650P Proxy | TFLITE | 46.425 ms | 4 - 51 MB | FP16 | NPU | [FastSam-X.tflite](https://huggingface.co/qualcomm/FastSam-X/blob/main/FastSam-X.tflite) | | FastSam-X | SA8650 (Proxy) | SA8650P Proxy | QNN | 43.452 ms | 5 - 7 MB | FP16 | NPU | Use Export Script | | FastSam-X | SA8775P ADP | SA8775P | TFLITE | 87.589 ms | 0 - 57 MB | FP16 | NPU | [FastSam-X.tflite](https://huggingface.co/qualcomm/FastSam-X/blob/main/FastSam-X.tflite) | | FastSam-X | SA8775P ADP | SA8775P | QNN | 85.645 ms | 2 - 12 MB | FP16 | NPU | Use Export Script | | FastSam-X | QCS8450 (Proxy) | QCS8450 Proxy | TFLITE | 90.22 ms | 4 - 63 MB | FP16 | NPU | [FastSam-X.tflite](https://huggingface.co/qualcomm/FastSam-X/blob/main/FastSam-X.tflite) | | FastSam-X | QCS8450 (Proxy) | QCS8450 Proxy | QNN | 88.072 ms | 0 - 59 MB | FP16 | NPU | Use Export Script | | FastSam-X | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN | 44.542 ms | 5 - 5 MB | FP16 | NPU | Use Export Script | | FastSam-X | Snapdragon X Elite CRD | Snapdragon® X Elite | ONNX | 49.563 ms | 141 - 141 MB | FP16 | NPU | [FastSam-X.onnx](https://huggingface.co/qualcomm/FastSam-X/blob/main/FastSam-X.onnx) | ## Installation Install the package via pip: ```bash pip install "qai-hub-models[fastsam-x]" ``` ## 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.fastsam_x.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.fastsam_x.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.fastsam_x.export ``` ``` Profiling Results ------------------------------------------------------------ FastSam-X Device : Samsung Galaxy S23 (13) Runtime : TFLITE Estimated inference time (ms) : 46.3 Estimated peak memory usage (MB): [4, 19] Total # Ops : 419 Compute Unit(s) : NPU (419 ops) ``` ## How does this work? This [export script](https://aihub.qualcomm.com/models/fastsam_x/qai_hub_models/models/FastSam-X/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.fastsam_x 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.fastsam_x.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.fastsam_x.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 FastSam-X's performance across various devices [here](https://aihub.qualcomm.com/models/fastsam_x). Explore all available models on [Qualcomm® AI Hub](https://aihub.qualcomm.com/) ## License * The license for the original implementation of FastSam-X can be found [here](https://github.com/CASIA-IVA-Lab/FastSAM/blob/main/LICENSE). * The license for the compiled assets for on-device deployment can be found [here](https://github.com/CASIA-IVA-Lab/FastSAM/blob/main/LICENSE) ## References * [Fast Segment Anything](https://arxiv.org/abs/2306.12156) * [Source Model Implementation](https://github.com/CASIA-IVA-Lab/FastSAM) ## 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).