Segment-Anything-Model-2: Optimized for Qualcomm Devices

SAM 2, the successor to Meta's Segment Anything Model (SAM), is a cutting-edge tool designed for comprehensive object segmentation in both images and videos. It excels in handling complex visual data through a unified, promptable model architecture that supports real-time processing and zero-shot generalization.

This is based on the implementation of Segment-Anything-Model-2 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 2.42, ONNX Runtime 1.24.3	Download
QNN_DLC	float	Universal	QAIRT 2.45	Download
QNN_DLC	w8a8	Universal	QAIRT 2.45	Download

For more device-specific assets and performance metrics, visit Segment-Anything-Model-2 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 Segment-Anything-Model-2 on GitHub for usage instructions.

Model Details

Model Type: Model_use_case.semantic_segmentation

Model Stats:

• Model checkpoint: sam2.1_hiera_t
• Input resolution: 720p (720x1280)
• Number of parameters (encoder): 33.5M
• Model size (encoder) (float): 128 MB
• Number of parameters (decoder): 6.22M
• Model size (decoder) (float): 23.7 MB

Performance Summary

Model	Runtime	Precision	Chipset	Inference Time (ms)	Peak Memory Range (MB)	Primary Compute Unit
decoder	ONNX	float	Snapdragon® 8 Elite Gen 5 Mobile	3.149 ms	17 - 242 MB	NPU
decoder	ONNX	float	Snapdragon® X2 Elite	3.22 ms	15 - 15 MB	NPU
decoder	ONNX	float	Snapdragon® 8 Gen 3 Mobile	4.729 ms	16 - 328 MB	NPU
decoder	ONNX	float	Qualcomm® QCS8550 (Proxy)	6.744 ms	16 - 24 MB	NPU
decoder	ONNX	float	Qualcomm® QCS9075	8.485 ms	16 - 19 MB	NPU
decoder	ONNX	float	Snapdragon® 8 Elite For Galaxy Mobile	3.498 ms	3 - 235 MB	NPU
decoder	QNN_DLC	float	Snapdragon® X2 Elite	3.057 ms	16 - 16 MB	NPU
decoder	QNN_DLC	float	Snapdragon® X Elite	5.549 ms	16 - 16 MB	NPU
decoder	QNN_DLC	float	Snapdragon® 8 Gen 3 Mobile	3.689 ms	4 - 213 MB	NPU
decoder	QNN_DLC	float	Qualcomm® QCS8550 (Proxy)	5.358 ms	16 - 18 MB	NPU
decoder	QNN_DLC	float	Qualcomm® QCS9075	6.826 ms	16 - 34 MB	NPU
decoder	QNN_DLC	float	Qualcomm® SA8295P	8.19 ms	0 - 189 MB	NPU
decoder	QNN_DLC	float	Snapdragon® 8 Elite For Galaxy Mobile	2.861 ms	4 - 186 MB	NPU
decoder	QNN_DLC	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	0.9 ms	0 - 126 MB	NPU
decoder	QNN_DLC	w8a8	Snapdragon® X2 Elite	1.334 ms	4 - 4 MB	NPU
decoder	QNN_DLC	w8a8	Snapdragon® X Elite	2.302 ms	4 - 4 MB	NPU
decoder	QNN_DLC	w8a8	Snapdragon® 8 Gen 3 Mobile	1.288 ms	0 - 107 MB	NPU
decoder	QNN_DLC	w8a8	Qualcomm® QCS6490	8.445 ms	1 - 7 MB	NPU
decoder	QNN_DLC	w8a8	Qualcomm® QCS8275 (Proxy)	3.935 ms	4 - 88 MB	NPU
decoder	QNN_DLC	w8a8	Qualcomm® QCS8550 (Proxy)	1.976 ms	4 - 6 MB	NPU
decoder	QNN_DLC	w8a8	Qualcomm® SA8775P	2.666 ms	4 - 91 MB	NPU
decoder	QNN_DLC	w8a8	Qualcomm® QCS9075	2.415 ms	1 - 7 MB	NPU
decoder	QNN_DLC	w8a8	Qualcomm® QCM6690	9.958 ms	4 - 270 MB	NPU
decoder	QNN_DLC	w8a8	Qualcomm® QCS8450 (Proxy)	2.561 ms	0 - 110 MB	NPU
decoder	QNN_DLC	w8a8	Qualcomm® SA7255P	3.935 ms	4 - 88 MB	NPU
decoder	QNN_DLC	w8a8	Qualcomm® SA8295P	2.939 ms	4 - 87 MB	NPU
decoder	QNN_DLC	w8a8	Snapdragon® 8 Elite For Galaxy Mobile	1.013 ms	1 - 90 MB	NPU
decoder	QNN_DLC	w8a8	Snapdragon® 7 Gen 4 Mobile	2.732 ms	4 - 211 MB	NPU
encoder	ONNX	float	Snapdragon® 8 Elite Gen 5 Mobile	98.074 ms	0 - 2232 MB	NPU
encoder	ONNX	float	Snapdragon® X2 Elite	99.93 ms	71 - 71 MB	NPU
encoder	ONNX	float	Snapdragon® 8 Gen 3 Mobile	143.51 ms	23 - 3816 MB	NPU
encoder	ONNX	float	Qualcomm® QCS8550 (Proxy)	197.238 ms	36 - 43 MB	NPU
encoder	ONNX	float	Qualcomm® QCS9075	272.475 ms	12 - 27 MB	NPU
encoder	ONNX	float	Snapdragon® 8 Elite For Galaxy Mobile	113.634 ms	33 - 1687 MB	NPU
encoder	QNN_DLC	float	Snapdragon® X2 Elite	108.94 ms	12 - 12 MB	NPU
encoder	QNN_DLC	float	Snapdragon® X Elite	206.568 ms	12 - 12 MB	NPU
encoder	QNN_DLC	float	Snapdragon® 8 Gen 3 Mobile	156.256 ms	29 - 1987 MB	NPU
encoder	QNN_DLC	float	Qualcomm® QCS8550 (Proxy)	207.936 ms	12 - 15 MB	NPU
encoder	QNN_DLC	float	Qualcomm® QCS9075	258.802 ms	12 - 42 MB	NPU
encoder	QNN_DLC	float	Qualcomm® SA8295P	309.402 ms	0 - 1671 MB	NPU
encoder	QNN_DLC	float	Snapdragon® 8 Elite For Galaxy Mobile	121.114 ms	12 - 1747 MB	NPU
encoder	QNN_DLC	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	38.857 ms	3 - 1809 MB	NPU
encoder	QNN_DLC	w8a8	Snapdragon® X2 Elite	40.756 ms	3 - 3 MB	NPU
encoder	QNN_DLC	w8a8	Snapdragon® X Elite	84.928 ms	3 - 3 MB	NPU
encoder	QNN_DLC	w8a8	Snapdragon® 8 Gen 3 Mobile	57.345 ms	3 - 1981 MB	NPU
encoder	QNN_DLC	w8a8	Qualcomm® QCS6490	665.154 ms	3 - 12 MB	NPU
encoder	QNN_DLC	w8a8	Qualcomm® QCS8275 (Proxy)	144.989 ms	3 - 1664 MB	NPU
encoder	QNN_DLC	w8a8	Qualcomm® QCS8550 (Proxy)	82.026 ms	3 - 6 MB	NPU
encoder	QNN_DLC	w8a8	Qualcomm® SA8775P	83.967 ms	3 - 1664 MB	NPU
encoder	QNN_DLC	w8a8	Qualcomm® QCS9075	103.371 ms	3 - 12 MB	NPU
encoder	QNN_DLC	w8a8	Qualcomm® QCM6690	1564.069 ms	3 - 1968 MB	NPU
encoder	QNN_DLC	w8a8	Qualcomm® QCS8450 (Proxy)	140.201 ms	3 - 2037 MB	NPU
encoder	QNN_DLC	w8a8	Qualcomm® SA7255P	144.989 ms	3 - 1664 MB	NPU
encoder	QNN_DLC	w8a8	Qualcomm® SA8295P	97.146 ms	3 - 1671 MB	NPU
encoder	QNN_DLC	w8a8	Snapdragon® 8 Elite For Galaxy Mobile	42.779 ms	3 - 1736 MB	NPU
encoder	QNN_DLC	w8a8	Snapdragon® 7 Gen 4 Mobile	110.817 ms	3 - 1775 MB	NPU

License

• The license for the original implementation of Segment-Anything-Model-2 can be found here.

References

• SAM 2 Segment Anything in Images and Videos
• 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.