FastSam-X: 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-X 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-X 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-X on GitHub for usage instructions.

Model Details

Model Type: Model_use_case.semantic_segmentation

Model Stats:

• Model checkpoint: fastsam-x.pt
• Inference latency: RealTime
• Input resolution: 640x640
• Number of parameters: 72.2M
• Model size (float): 276 MB

Performance Summary

Model	Runtime	Precision	Chipset	Inference Time (ms)	Peak Memory Range (MB)	Primary Compute Unit
FastSam-X	ONNX	float	Snapdragon® X2 Elite	24.41 ms	17 - 17 MB	NPU
FastSam-X	ONNX	float	Snapdragon® X Elite	47.023 ms	139 - 139 MB	NPU
FastSam-X	ONNX	float	Snapdragon® 8 Gen 3 Mobile	36.186 ms	2 - 320 MB	NPU
FastSam-X	ONNX	float	Snapdragon® 8 Gen 1 Mobile	97.029 ms	17 - 415 MB	NPU
FastSam-X	ONNX	float	Qualcomm® Dragonwing™ QCS8550 (Proxy)	48.161 ms	0 - 169 MB	NPU
FastSam-X	ONNX	float	Qualcomm® QCS8450	97.029 ms	17 - 415 MB	NPU
FastSam-X	ONNX	float	Qualcomm® Dragonwing™ IQ-9075	72.529 ms	12 - 19 MB	NPU
FastSam-X	ONNX	float	Qualcomm® Dragonwing™ IQ-X7181	47.023 ms	139 - 139 MB	NPU
FastSam-X	ONNX	float	Qualcomm® Dragonwing™ Q-8750	27.533 ms	13 - 244 MB	NPU
FastSam-X	ONNX	float	Snapdragon® 8 Elite Mobile	27.533 ms	13 - 244 MB	NPU
FastSam-X	ONNX	float	Snapdragon® 8 Elite Gen 5 Mobile	19.891 ms	12 - 266 MB	NPU
FastSam-X	ONNX	w8a16	Snapdragon® X2 Elite	15.635 ms	9 - 9 MB	NPU
FastSam-X	ONNX	w8a16	Snapdragon® X Elite	33.634 ms	72 - 72 MB	NPU
FastSam-X	ONNX	w8a16	Snapdragon® 8 Gen 3 Mobile	24.748 ms	0 - 453 MB	NPU
FastSam-X	ONNX	w8a16	Snapdragon® 8 Gen 1 Mobile	58.179 ms	1 - 455 MB	NPU
FastSam-X	ONNX	w8a16	Qualcomm® Dragonwing™ QCS6490	183.988 ms	6 - 11 MB	NPU
FastSam-X	ONNX	w8a16	Qualcomm® Dragonwing™ QCS8550 (Proxy)	34.28 ms	0 - 81 MB	NPU
FastSam-X	ONNX	w8a16	Qualcomm® QCS8450	58.179 ms	1 - 455 MB	NPU
FastSam-X	ONNX	w8a16	Qualcomm® Dragonwing™ IQ-9075	33.613 ms	6 - 11 MB	NPU
FastSam-X	ONNX	w8a16	Qualcomm® Dragonwing™ IQ-X7181	33.634 ms	72 - 72 MB	NPU
FastSam-X	ONNX	w8a16	Qualcomm® Dragonwing™ Q-8750	19.989 ms	6 - 364 MB	NPU
FastSam-X	ONNX	w8a16	Snapdragon® 8 Elite Mobile	19.989 ms	6 - 364 MB	NPU
FastSam-X	ONNX	w8a16	Snapdragon® 8 Elite Gen 5 Mobile	12.897 ms	6 - 370 MB	NPU
FastSam-X	QNN_DLC	float	Snapdragon® X2 Elite	23.182 ms	5 - 5 MB	NPU
FastSam-X	QNN_DLC	float	Snapdragon® X Elite	44.046 ms	5 - 5 MB	NPU
FastSam-X	QNN_DLC	float	Snapdragon® 8 Gen 3 Mobile	32.837 ms	5 - 302 MB	NPU
FastSam-X	QNN_DLC	float	Snapdragon® 8 Gen 1 Mobile	92.096 ms	2 - 384 MB	NPU
FastSam-X	QNN_DLC	float	Qualcomm® Dragonwing™ QCS8275	279.847 ms	0 - 215 MB	NPU
FastSam-X	QNN_DLC	float	Qualcomm® Dragonwing™ QCS8550 (Proxy)	43.111 ms	5 - 7 MB	NPU
FastSam-X	QNN_DLC	float	Qualcomm® SA8775P	68.48 ms	1 - 214 MB	NPU
FastSam-X	QNN_DLC	float	Qualcomm® SA8650P	68.48 ms	1 - 214 MB	NPU
FastSam-X	QNN_DLC	float	Qualcomm® SA8255P	68.48 ms	1 - 214 MB	NPU
FastSam-X	QNN_DLC	float	Qualcomm® QCS8450	92.096 ms	2 - 384 MB	NPU
FastSam-X	QNN_DLC	float	Qualcomm® Dragonwing™ IQ-9075	71.034 ms	5 - 15 MB	NPU
FastSam-X	QNN_DLC	float	Qualcomm® Dragonwing™ IQ-X7181	44.046 ms	5 - 5 MB	NPU
FastSam-X	QNN_DLC	float	Qualcomm® Dragonwing™ Q-8750	25.399 ms	5 - 225 MB	NPU
FastSam-X	QNN_DLC	float	Qualcomm® SA7255P	279.847 ms	0 - 215 MB	NPU
FastSam-X	QNN_DLC	float	Qualcomm® SA8295P	77.696 ms	0 - 298 MB	NPU
FastSam-X	QNN_DLC	float	Snapdragon® 8 Elite Mobile	25.399 ms	5 - 225 MB	NPU
FastSam-X	QNN_DLC	float	Snapdragon® 8 Elite Gen 5 Mobile	17.976 ms	5 - 238 MB	NPU
FastSam-X	QNN_DLC	w8a16	Snapdragon® X2 Elite	15.167 ms	2 - 2 MB	NPU
FastSam-X	QNN_DLC	w8a16	Snapdragon® X Elite	31.422 ms	2 - 2 MB	NPU
FastSam-X	QNN_DLC	w8a16	Snapdragon® 8 Gen 3 Mobile	22.331 ms	2 - 421 MB	NPU
FastSam-X	QNN_DLC	w8a16	Snapdragon® 8 Gen 1 Mobile	50.538 ms	2 - 416 MB	NPU
FastSam-X	QNN_DLC	w8a16	Qualcomm® Dragonwing™ QCS6490	179.392 ms	3 - 10 MB	NPU
FastSam-X	QNN_DLC	w8a16	Qualcomm® Dragonwing™ QCS8275	86.274 ms	0 - 319 MB	NPU
FastSam-X	QNN_DLC	w8a16	Qualcomm® Dragonwing™ QCS8550 (Proxy)	31.238 ms	3 - 5 MB	NPU
FastSam-X	QNN_DLC	w8a16	Qualcomm® SA8775P	28.463 ms	1 - 323 MB	NPU
FastSam-X	QNN_DLC	w8a16	Qualcomm® SA8650P	28.463 ms	1 - 323 MB	NPU
FastSam-X	QNN_DLC	w8a16	Qualcomm® SA8255P	28.463 ms	1 - 323 MB	NPU
FastSam-X	QNN_DLC	w8a16	Qualcomm® QCS8450	50.538 ms	2 - 416 MB	NPU
FastSam-X	QNN_DLC	w8a16	Qualcomm® Dragonwing™ IQ-9075	28.79 ms	2 - 8 MB	NPU
FastSam-X	QNN_DLC	w8a16	Qualcomm® Dragonwing™ IQ-X7181	31.422 ms	2 - 2 MB	NPU
FastSam-X	QNN_DLC	w8a16	Qualcomm® Dragonwing™ Q-6690	690.956 ms	3 - 480 MB	NPU
FastSam-X	QNN_DLC	w8a16	Qualcomm® Dragonwing™ Q-7790	53.458 ms	2 - 398 MB	NPU
FastSam-X	QNN_DLC	w8a16	Qualcomm® Dragonwing™ Q-8750	18.568 ms	2 - 331 MB	NPU
FastSam-X	QNN_DLC	w8a16	Qualcomm® SA7255P	86.274 ms	0 - 319 MB	NPU
FastSam-X	QNN_DLC	w8a16	Qualcomm® SA8295P	45.085 ms	0 - 318 MB	NPU
FastSam-X	QNN_DLC	w8a16	Snapdragon® 8 Elite Mobile	18.568 ms	2 - 331 MB	NPU
FastSam-X	QNN_DLC	w8a16	Snapdragon® 8 Elite Gen 5 Mobile	12.009 ms	2 - 333 MB	NPU
FastSam-X	QNN_DLC	w8a16	Snapdragon® 7 Gen 4 Mobile	53.458 ms	2 - 398 MB	NPU
FastSam-X	TFLITE	float	Snapdragon® 8 Gen 3 Mobile	32.398 ms	4 - 418 MB	NPU
FastSam-X	TFLITE	float	Snapdragon® 8 Gen 1 Mobile	95.151 ms	5 - 506 MB	NPU
FastSam-X	TFLITE	float	Qualcomm® Dragonwing™ QCS8275	279.375 ms	5 - 257 MB	NPU
FastSam-X	TFLITE	float	Qualcomm® Dragonwing™ QCS8550 (Proxy)	43.623 ms	5 - 8 MB	NPU
FastSam-X	TFLITE	float	Qualcomm® SA8775P	67.958 ms	5 - 257 MB	NPU
FastSam-X	TFLITE	float	Qualcomm® SA8650P	67.958 ms	5 - 257 MB	NPU
FastSam-X	TFLITE	float	Qualcomm® SA8255P	67.958 ms	5 - 257 MB	NPU
FastSam-X	TFLITE	float	Qualcomm® QCS8450	95.151 ms	5 - 506 MB	NPU
FastSam-X	TFLITE	float	Qualcomm® Dragonwing™ IQ-9075	68.418 ms	3 - 156 MB	NPU
FastSam-X	TFLITE	float	Qualcomm® Dragonwing™ Q-8750	25.634 ms	3 - 260 MB	NPU
FastSam-X	TFLITE	float	Qualcomm® SA7255P	279.375 ms	5 - 257 MB	NPU
FastSam-X	TFLITE	float	Qualcomm® SA8295P	77.128 ms	5 - 337 MB	NPU
FastSam-X	TFLITE	float	Snapdragon® 8 Elite Mobile	25.634 ms	3 - 260 MB	NPU
FastSam-X	TFLITE	float	Snapdragon® 8 Elite Gen 5 Mobile	17.447 ms	2 - 266 MB	NPU

License

• The license for the original implementation of FastSam-X 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.