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 2.37, ONNX Runtime 1.23.0	Download
QNN_DLC	float	Universal	QAIRT 2.42	Download
TFLITE	float	Universal	QAIRT 2.42, TFLite 2.17.0	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® X Elite	8.598 ms	20 - 20 MB	NPU
FastSam-S	ONNX	float	Snapdragon® 8 Gen 3 Mobile	6.1 ms	22 - 224 MB	NPU
FastSam-S	ONNX	float	Qualcomm® QCS8550 (Proxy)	8.371 ms	0 - 38 MB	NPU
FastSam-S	ONNX	float	Qualcomm® QCS9075	13.529 ms	12 - 15 MB	NPU
FastSam-S	ONNX	float	Snapdragon® 8 Elite For Galaxy Mobile	4.789 ms	12 - 177 MB	NPU
FastSam-S	ONNX	float	Snapdragon® 8 Elite Gen 5 Mobile	3.561 ms	2 - 150 MB	NPU
FastSam-S	QNN_DLC	float	Snapdragon® X Elite	8.011 ms	5 - 5 MB	NPU
FastSam-S	QNN_DLC	float	Snapdragon® 8 Gen 3 Mobile	5.621 ms	0 - 255 MB	NPU
FastSam-S	QNN_DLC	float	Qualcomm® QCS8275 (Proxy)	38.556 ms	1 - 200 MB	NPU
FastSam-S	QNN_DLC	float	Qualcomm® QCS8550 (Proxy)	7.446 ms	5 - 10 MB	NPU
FastSam-S	QNN_DLC	float	Qualcomm® SA8775P	11.282 ms	1 - 207 MB	NPU
FastSam-S	QNN_DLC	float	Qualcomm® QCS9075	10.872 ms	7 - 17 MB	NPU
FastSam-S	QNN_DLC	float	Qualcomm® QCS8450 (Proxy)	15.055 ms	5 - 212 MB	NPU
FastSam-S	QNN_DLC	float	Qualcomm® SA7255P	38.556 ms	1 - 200 MB	NPU
FastSam-S	QNN_DLC	float	Qualcomm® SA8295P	13.778 ms	0 - 176 MB	NPU
FastSam-S	QNN_DLC	float	Snapdragon® 8 Elite For Galaxy Mobile	4.334 ms	0 - 202 MB	NPU
FastSam-S	QNN_DLC	float	Snapdragon® 8 Elite Gen 5 Mobile	3.137 ms	5 - 201 MB	NPU
FastSam-S	TFLITE	float	Snapdragon® 8 Gen 3 Mobile	5.159 ms	3 - 174 MB	NPU
FastSam-S	TFLITE	float	Qualcomm® QCS8275 (Proxy)	37.71 ms	4 - 115 MB	NPU
FastSam-S	TFLITE	float	Qualcomm® QCS8550 (Proxy)	6.862 ms	4 - 33 MB	NPU
FastSam-S	TFLITE	float	Qualcomm® SA8775P	10.644 ms	4 - 120 MB	NPU
FastSam-S	TFLITE	float	Qualcomm® QCS9075	10.565 ms	4 - 39 MB	NPU
FastSam-S	TFLITE	float	Qualcomm® QCS8450 (Proxy)	14.079 ms	4 - 231 MB	NPU
FastSam-S	TFLITE	float	Qualcomm® SA7255P	37.71 ms	4 - 115 MB	NPU
FastSam-S	TFLITE	float	Qualcomm® SA8295P	12.955 ms	4 - 195 MB	NPU
FastSam-S	TFLITE	float	Snapdragon® 8 Elite For Galaxy Mobile	3.893 ms	0 - 111 MB	NPU
FastSam-S	TFLITE	float	Snapdragon® 8 Elite Gen 5 Mobile	2.881 ms	0 - 208 MB	NPU

License

• The license for the original implementation of FastSam-S can be found here.

References

• Fast Segment Anything
• Source Model Implementation

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.