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 1.27.1	Download
ONNX	w8a16	Universal	QAIRT 2.45, ONNX Runtime 1.27.1	Download
QNN_DLC	float	Universal	QAIRT 2.45	Download
QNN_DLC	w8a16	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	4.554 ms	17 - 17 MB	NPU
FastSam-S	ONNX	float	Snapdragon® X Elite	8.527 ms	21 - 21 MB	NPU
FastSam-S	ONNX	float	Snapdragon® 8 Gen 3 Mobile	5.997 ms	17 - 244 MB	NPU
FastSam-S	ONNX	float	Snapdragon® 8 Gen 1 Mobile	16.539 ms	17 - 237 MB	NPU
FastSam-S	ONNX	float	Qualcomm® Dragonwing™ QCS8550 (Proxy)	8.194 ms	0 - 44 MB	NPU
FastSam-S	ONNX	float	Qualcomm® QCS8450	16.539 ms	17 - 237 MB	NPU
FastSam-S	ONNX	float	Qualcomm® Dragonwing™ IQ-9075	12.01 ms	15 - 18 MB	NPU
FastSam-S	ONNX	float	Qualcomm® Dragonwing™ IQ-X7181	8.527 ms	21 - 21 MB	NPU
FastSam-S	ONNX	float	Qualcomm® Dragonwing™ Q-8750	4.692 ms	13 - 213 MB	NPU
FastSam-S	ONNX	float	Snapdragon® 8 Elite Mobile	4.692 ms	13 - 213 MB	NPU
FastSam-S	ONNX	float	Snapdragon® 8 Elite Gen 5 Mobile	3.642 ms	1 - 208 MB	NPU
FastSam-S	ONNX	w8a16	Snapdragon® X2 Elite	3.119 ms	9 - 9 MB	NPU
FastSam-S	ONNX	w8a16	Snapdragon® X Elite	7.08 ms	10 - 10 MB	NPU
FastSam-S	ONNX	w8a16	Snapdragon® 8 Gen 3 Mobile	4.404 ms	8 - 264 MB	NPU
FastSam-S	ONNX	w8a16	Snapdragon® 8 Gen 1 Mobile	11.446 ms	8 - 266 MB	NPU
FastSam-S	ONNX	w8a16	Qualcomm® Dragonwing™ QCS6490	28.225 ms	8 - 10 MB	NPU
FastSam-S	ONNX	w8a16	Qualcomm® Dragonwing™ QCS8550 (Proxy)	6.675 ms	0 - 190 MB	NPU
FastSam-S	ONNX	w8a16	Qualcomm® QCS8450	11.446 ms	8 - 266 MB	NPU
FastSam-S	ONNX	w8a16	Qualcomm® Dragonwing™ IQ-9075	7.007 ms	8 - 10 MB	NPU
FastSam-S	ONNX	w8a16	Qualcomm® Dragonwing™ IQ-X7181	7.08 ms	10 - 10 MB	NPU
FastSam-S	ONNX	w8a16	Qualcomm® Dragonwing™ Q-8750	3.279 ms	0 - 220 MB	NPU
FastSam-S	ONNX	w8a16	Snapdragon® 8 Elite Mobile	3.279 ms	0 - 220 MB	NPU
FastSam-S	ONNX	w8a16	Snapdragon® 8 Elite Gen 5 Mobile	2.69 ms	0 - 225 MB	NPU
FastSam-S	QNN_DLC	float	Snapdragon® X2 Elite	4.593 ms	5 - 5 MB	NPU
FastSam-S	QNN_DLC	float	Snapdragon® X Elite	8.423 ms	5 - 5 MB	NPU
FastSam-S	QNN_DLC	float	Snapdragon® 8 Gen 3 Mobile	5.85 ms	5 - 216 MB	NPU
FastSam-S	QNN_DLC	float	Snapdragon® 8 Gen 1 Mobile	16.075 ms	5 - 213 MB	NPU
FastSam-S	QNN_DLC	float	Qualcomm® Dragonwing™ QCS8275	39.17 ms	1 - 187 MB	NPU
FastSam-S	QNN_DLC	float	Qualcomm® Dragonwing™ QCS8550 (Proxy)	7.706 ms	5 - 133 MB	NPU
FastSam-S	QNN_DLC	float	Qualcomm® SA8775P	11.584 ms	2 - 184 MB	NPU
FastSam-S	QNN_DLC	float	Qualcomm® SA8650P	11.584 ms	2 - 184 MB	NPU
FastSam-S	QNN_DLC	float	Qualcomm® SA8255P	11.584 ms	2 - 184 MB	NPU
FastSam-S	QNN_DLC	float	Qualcomm® QCS8450	16.075 ms	5 - 213 MB	NPU
FastSam-S	QNN_DLC	float	Qualcomm® Dragonwing™ IQ-9075	11.454 ms	7 - 17 MB	NPU
FastSam-S	QNN_DLC	float	Qualcomm® Dragonwing™ IQ-X7181	8.423 ms	5 - 5 MB	NPU
FastSam-S	QNN_DLC	float	Qualcomm® Dragonwing™ Q-8750	4.488 ms	5 - 195 MB	NPU
FastSam-S	QNN_DLC	float	Qualcomm® SA7255P	39.17 ms	1 - 187 MB	NPU
FastSam-S	QNN_DLC	float	Qualcomm® SA8295P	13.795 ms	0 - 177 MB	NPU
FastSam-S	QNN_DLC	float	Snapdragon® 8 Elite Mobile	4.488 ms	5 - 195 MB	NPU
FastSam-S	QNN_DLC	float	Snapdragon® 8 Elite Gen 5 Mobile	3.255 ms	5 - 198 MB	NPU
FastSam-S	QNN_DLC	w8a16	Snapdragon® X2 Elite	2.915 ms	2 - 2 MB	NPU
FastSam-S	QNN_DLC	w8a16	Snapdragon® X Elite	6.052 ms	2 - 2 MB	NPU
FastSam-S	QNN_DLC	w8a16	Snapdragon® 8 Gen 3 Mobile	3.782 ms	2 - 233 MB	NPU
FastSam-S	QNN_DLC	w8a16	Snapdragon® 8 Gen 1 Mobile	8.316 ms	2 - 233 MB	NPU
FastSam-S	QNN_DLC	w8a16	Qualcomm® Dragonwing™ QCS6490	23.921 ms	2 - 8 MB	NPU
FastSam-S	QNN_DLC	w8a16	Qualcomm® Dragonwing™ QCS8275	13.375 ms	1 - 195 MB	NPU
FastSam-S	QNN_DLC	w8a16	Qualcomm® Dragonwing™ QCS8550 (Proxy)	5.457 ms	2 - 4 MB	NPU
FastSam-S	QNN_DLC	w8a16	Qualcomm® SA8775P	5.992 ms	1 - 197 MB	NPU
FastSam-S	QNN_DLC	w8a16	Qualcomm® SA8650P	5.992 ms	1 - 197 MB	NPU
FastSam-S	QNN_DLC	w8a16	Qualcomm® SA8255P	5.992 ms	1 - 197 MB	NPU
FastSam-S	QNN_DLC	w8a16	Qualcomm® QCS8450	8.316 ms	2 - 233 MB	NPU
FastSam-S	QNN_DLC	w8a16	Qualcomm® Dragonwing™ IQ-9075	5.761 ms	2 - 8 MB	NPU
FastSam-S	QNN_DLC	w8a16	Qualcomm® Dragonwing™ IQ-X7181	6.052 ms	2 - 2 MB	NPU
FastSam-S	QNN_DLC	w8a16	Qualcomm® Dragonwing™ Q-6690	66.101 ms	2 - 204 MB	NPU
FastSam-S	QNN_DLC	w8a16	Qualcomm® Dragonwing™ Q-7790	7.769 ms	2 - 195 MB	NPU
FastSam-S	QNN_DLC	w8a16	Qualcomm® Dragonwing™ Q-8750	2.675 ms	2 - 200 MB	NPU
FastSam-S	QNN_DLC	w8a16	Qualcomm® SA7255P	13.375 ms	1 - 195 MB	NPU
FastSam-S	QNN_DLC	w8a16	Qualcomm® SA8295P	7.881 ms	0 - 191 MB	NPU
FastSam-S	QNN_DLC	w8a16	Snapdragon® 8 Elite Mobile	2.675 ms	2 - 200 MB	NPU
FastSam-S	QNN_DLC	w8a16	Snapdragon® 8 Elite Gen 5 Mobile	2.11 ms	2 - 204 MB	NPU
FastSam-S	QNN_DLC	w8a16	Snapdragon® 7 Gen 4 Mobile	7.769 ms	2 - 195 MB	NPU
FastSam-S	TFLITE	float	Snapdragon® 8 Gen 3 Mobile	5.172 ms	4 - 117 MB	NPU
FastSam-S	TFLITE	float	Snapdragon® 8 Gen 1 Mobile	14.211 ms	0 - 228 MB	NPU
FastSam-S	TFLITE	float	Qualcomm® Dragonwing™ QCS8275	37.805 ms	4 - 86 MB	NPU
FastSam-S	TFLITE	float	Qualcomm® Dragonwing™ QCS8550 (Proxy)	6.949 ms	4 - 6 MB	NPU
FastSam-S	TFLITE	float	Qualcomm® SA8775P	10.732 ms	4 - 90 MB	NPU
FastSam-S	TFLITE	float	Qualcomm® SA8650P	10.732 ms	4 - 90 MB	NPU
FastSam-S	TFLITE	float	Qualcomm® SA8255P	10.732 ms	4 - 90 MB	NPU
FastSam-S	TFLITE	float	Qualcomm® QCS8450	14.211 ms	0 - 228 MB	NPU
FastSam-S	TFLITE	float	Qualcomm® Dragonwing™ IQ-9075	10.413 ms	4 - 39 MB	NPU
FastSam-S	TFLITE	float	Qualcomm® Dragonwing™ Q-8750	3.947 ms	3 - 92 MB	NPU
FastSam-S	TFLITE	float	Qualcomm® SA7255P	37.805 ms	4 - 86 MB	NPU
FastSam-S	TFLITE	float	Qualcomm® SA8295P	13.094 ms	4 - 195 MB	NPU
FastSam-S	TFLITE	float	Snapdragon® 8 Elite Mobile	3.947 ms	3 - 92 MB	NPU
FastSam-S	TFLITE	float	Snapdragon® 8 Elite Gen 5 Mobile	3.124 ms	0 - 96 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.