Fast Segment Anything
Paper • 2306.12156 • Published • 36
FastSam-S: Optimized for Qualcomm Devices
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 is based on the implementation of FastSam-S found here. This repository contains pre-exported model files optimized for Qualcomm® devices. You can use the Qualcomm® AI Hub Models library to export with custom configurations. More details on model performance across various devices, can be found here.
Qualcomm AI Hub Models uses Qualcomm AI Hub Workbench to compile, profile, and evaluate this model. Sign up to run these models on a hosted Qualcomm® device.
Getting Started
There are two ways to deploy this model on your device:
Option 1: Download Pre-Exported Models
Below are pre-exported model assets ready for deployment.
| Runtime | Precision | Chipset | SDK Versions | Download |
|---|---|---|---|---|
| ONNX | float | Universal | QAIRT 2.45, ONNX Runtime 1.25.0 | Download |
| QNN_DLC | float | Universal | QAIRT 2.45 | Download |
| TFLITE | float | Universal | QAIRT 2.45 | Download |
For more device-specific assets and performance metrics, visit FastSam-S on Qualcomm® AI Hub.
Option 2: Export with Custom Configurations
Use the Qualcomm® AI Hub Models Python library to compile and export the model with your own:
- Custom weights (e.g., fine-tuned checkpoints)
- Custom input shapes
- Target device and runtime configurations
This option is ideal if you need to customize the model beyond the default configuration provided here.
See our repository for FastSam-S on GitHub for usage instructions.
Model Details
Model Type: Model_use_case.semantic_segmentation
Model Stats:
- Model checkpoint: fastsam-s.pt
- Inference latency: RealTime
- Input resolution: 640x640
- Number of parameters: 11.8M
- Model size (float): 45.1 MB
Performance Summary
| Model | Runtime | Precision | Chipset | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit |
|---|---|---|---|---|---|---|
| FastSam-S | ONNX | float | Snapdragon® X2 Elite | 4.587 ms | 208 - 208 MB | NPU |
| FastSam-S | ONNX | float | Snapdragon® X Elite | 8.391 ms | 177 - 177 MB | NPU |
| FastSam-S | ONNX | float | Snapdragon® 8 Gen 3 Mobile | 5.994 ms | 2 - 220 MB | NPU |
| FastSam-S | ONNX | float | Snapdragon® 8 Gen 1 Mobile | 16.797 ms | 17 - 228 MB | NPU |
| FastSam-S | ONNX | float | Qualcomm® QCS8550 (Proxy) | 8.194 ms | 0 - 30 MB | NPU |
| FastSam-S | ONNX | float | Qualcomm® QCS8450 | 16.797 ms | 17 - 228 MB | NPU |
| FastSam-S | ONNX | float | Snapdragon® 8 Elite Mobile | 4.71 ms | 12 - 212 MB | NPU |
| FastSam-S | ONNX | float | Snapdragon® 8 Elite Gen 5 Mobile | 3.559 ms | 1 - 203 MB | NPU |
| FastSam-S | ONNX | float | Qualcomm® QCS9075 | 12.881 ms | 16 - 62 MB | NPU |
| FastSam-S | ONNX | float | Qualcomm® QCS8750 | 4.71 ms | 12 - 212 MB | NPU |
| FastSam-S | ONNX | float | Qualcomm® QCS7181 | 8.391 ms | 177 - 177 MB | NPU |
| FastSam-S | QNN_DLC | float | Snapdragon® X2 Elite | 4.574 ms | 5 - 5 MB | NPU |
| FastSam-S | QNN_DLC | float | Snapdragon® X Elite | 8.388 ms | 5 - 5 MB | NPU |
| FastSam-S | QNN_DLC | float | Snapdragon® 8 Gen 3 Mobile | 5.841 ms | 4 - 216 MB | NPU |
| FastSam-S | QNN_DLC | float | Snapdragon® 8 Gen 1 Mobile | 16.069 ms | 5 - 213 MB | NPU |
| FastSam-S | QNN_DLC | float | Qualcomm® QCS8275 | 39.211 ms | 1 - 187 MB | NPU |
| FastSam-S | QNN_DLC | float | Qualcomm® QCS8550 (Proxy) | 7.694 ms | 5 - 7 MB | NPU |
| FastSam-S | QNN_DLC | float | Qualcomm® QCS8450 | 16.069 ms | 5 - 213 MB | NPU |
| FastSam-S | QNN_DLC | float | Snapdragon® 8 Elite Mobile | 4.495 ms | 5 - 194 MB | NPU |
| FastSam-S | QNN_DLC | float | Qualcomm® SA8295P | 13.762 ms | 0 - 178 MB | NPU |
| FastSam-S | QNN_DLC | float | Snapdragon® 8 Elite Gen 5 Mobile | 3.169 ms | 5 - 199 MB | NPU |
| FastSam-S | QNN_DLC | float | Qualcomm® SA7255P | 39.211 ms | 1 - 187 MB | NPU |
| FastSam-S | QNN_DLC | float | Qualcomm® QCS9075 | 10.933 ms | 5 - 15 MB | NPU |
| FastSam-S | QNN_DLC | float | Qualcomm® QCS8750 | 4.495 ms | 5 - 194 MB | NPU |
| FastSam-S | QNN_DLC | float | Qualcomm® QCS7181 | 8.388 ms | 5 - 5 MB | NPU |
| FastSam-S | TFLITE | float | Snapdragon® 8 Gen 3 Mobile | 5.168 ms | 3 - 117 MB | NPU |
| FastSam-S | TFLITE | float | Snapdragon® 8 Gen 1 Mobile | 14.257 ms | 4 - 234 MB | NPU |
| FastSam-S | TFLITE | float | Qualcomm® QCS8275 | 37.809 ms | 4 - 86 MB | NPU |
| FastSam-S | TFLITE | float | Qualcomm® QCS8550 (Proxy) | 6.953 ms | 0 - 3 MB | NPU |
| FastSam-S | TFLITE | float | Qualcomm® SA8775P | 62.496 ms | 5 - 36 MB | GPU |
| FastSam-S | TFLITE | float | Qualcomm® SA8650P | 62.496 ms | 5 - 36 MB | GPU |
| FastSam-S | TFLITE | float | Qualcomm® SA8255P | 62.496 ms | 5 - 36 MB | GPU |
| FastSam-S | TFLITE | float | Qualcomm® QCS8450 | 14.257 ms | 4 - 234 MB | NPU |
| FastSam-S | TFLITE | float | Snapdragon® 8 Elite Mobile | 3.938 ms | 3 - 101 MB | NPU |
| FastSam-S | TFLITE | float | Qualcomm® SA8295P | 13.076 ms | 4 - 195 MB | NPU |
| FastSam-S | TFLITE | float | Snapdragon® 8 Elite Gen 5 Mobile | 3.03 ms | 0 - 91 MB | NPU |
| FastSam-S | TFLITE | float | Qualcomm® SA7255P | 37.809 ms | 4 - 86 MB | NPU |
| FastSam-S | TFLITE | float | Qualcomm® QCS9075 | 10.664 ms | 4 - 39 MB | NPU |
| FastSam-S | TFLITE | float | Qualcomm® QCS8750 | 3.938 ms | 3 - 101 MB | NPU |
License
- The license for the original implementation of FastSam-S 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.