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
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	3.153 ms	17 - 245 MB	NPU
decoder	ONNX	float	Snapdragon® 8 Gen 3 Mobile	4.754 ms	1 - 314 MB	NPU
decoder	ONNX	float	Qualcomm® QCS8550 (Proxy)	6.794 ms	16 - 23 MB	NPU
decoder	ONNX	float	Qualcomm® QCS9075	8.445 ms	16 - 19 MB	NPU
decoder	QNN_DLC	float	Snapdragon® 8 Elite Gen 5 Mobile	2.609 ms	4 - 190 MB	NPU
decoder	QNN_DLC	float	Snapdragon® 8 Elite Mobile	2.894 ms	4 - 186 MB	NPU
decoder	QNN_DLC	float	Snapdragon® X2 Elite	3.066 ms	16 - 16 MB	NPU
decoder	QNN_DLC	float	Snapdragon® X Elite	5.508 ms	16 - 16 MB	NPU
decoder	QNN_DLC	float	Snapdragon® X Elite	5.508 ms	16 - 16 MB	NPU
decoder	QNN_DLC	float	Snapdragon® 8 Gen 3 Mobile	3.724 ms	4 - 213 MB	NPU
decoder	QNN_DLC	float	Qualcomm® QCS8550 (Proxy)	5.328 ms	16 - 18 MB	NPU
decoder	QNN_DLC	float	Qualcomm® QCS8450 (Proxy)	10.001 ms	4 - 218 MB	NPU
decoder	QNN_DLC	float	Qualcomm® SA8295P	8.186 ms	1 - 194 MB	NPU
decoder	QNN_DLC	float	Qualcomm® QCS9075	6.743 ms	16 - 34 MB	NPU
decoder	QNN_DLC	float	Snapdragon® 8 Elite For Galaxy Mobile	2.894 ms	4 - 186 MB	NPU
decoder	QNN_DLC	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	0.89 ms	1 - 109 MB	NPU
decoder	QNN_DLC	w8a8	Snapdragon® 8 Elite Mobile	1.012 ms	1 - 90 MB	NPU
decoder	QNN_DLC	w8a8	Snapdragon® X2 Elite	1.36 ms	4 - 4 MB	NPU
decoder	QNN_DLC	w8a8	Snapdragon® X Elite	2.273 ms	4 - 4 MB	NPU
decoder	QNN_DLC	w8a8	Snapdragon® X Elite	2.273 ms	4 - 4 MB	NPU
decoder	QNN_DLC	w8a8	Snapdragon® 8 Gen 3 Mobile	1.292 ms	0 - 107 MB	NPU
decoder	QNN_DLC	w8a8	Qualcomm® QCS6490	8.961 ms	3 - 9 MB	NPU
decoder	QNN_DLC	w8a8	Qualcomm® QCS8550 (Proxy)	2.017 ms	4 - 19 MB	NPU
decoder	QNN_DLC	w8a8	Snapdragon® 8 Elite For Galaxy Mobile	1.012 ms	1 - 90 MB	NPU
decoder	QNN_DLC	w8a8	Qualcomm® QCS8450 (Proxy)	2.551 ms	0 - 109 MB	NPU
decoder	QNN_DLC	w8a8	Snapdragon® 7 Gen 4 Mobile	2.684 ms	4 - 214 MB	NPU
decoder	QNN_DLC	w8a8	Qualcomm® QCM6690	9.916 ms	4 - 272 MB	NPU
decoder	QNN_DLC	w8a8	Qualcomm® SA8295P	2.978 ms	4 - 87 MB	NPU
decoder	QNN_DLC	w8a8	Qualcomm® QCS9075	2.407 ms	0 - 6 MB	NPU
decoder	QNN_DLC	w8a8	Snapdragon® 7 Gen 4 Mobile	2.684 ms	4 - 214 MB	NPU
encoder	ONNX	float	Snapdragon® 8 Elite Gen 5 Mobile	79.657 ms	2 - 2223 MB	NPU
encoder	ONNX	float	Snapdragon® 8 Gen 3 Mobile	119.374 ms	1 - 1920 MB	NPU
encoder	ONNX	float	Qualcomm® QCS8550 (Proxy)	162.471 ms	0 - 99 MB	NPU
encoder	ONNX	float	Qualcomm® QCS9075	217.885 ms	12 - 27 MB	NPU
encoder	QNN_DLC	float	Snapdragon® 8 Elite Gen 5 Mobile	83.761 ms	12 - 2384 MB	NPU
encoder	QNN_DLC	float	Snapdragon® 8 Elite Mobile	97.831 ms	0 - 1690 MB	NPU
encoder	QNN_DLC	float	Snapdragon® X2 Elite	96.044 ms	12 - 12 MB	NPU
encoder	QNN_DLC	float	Snapdragon® X Elite	170.496 ms	12 - 12 MB	NPU
encoder	QNN_DLC	float	Snapdragon® X Elite	170.496 ms	12 - 12 MB	NPU
encoder	QNN_DLC	float	Snapdragon® 8 Gen 3 Mobile	134.146 ms	12 - 1955 MB	NPU
encoder	QNN_DLC	float	Qualcomm® QCS8550 (Proxy)	177.293 ms	8 - 14 MB	NPU
encoder	QNN_DLC	float	Qualcomm® QCS8450 (Proxy)	343.674 ms	13 - 1882 MB	NPU
encoder	QNN_DLC	float	Qualcomm® SA8295P	249.633 ms	0 - 1638 MB	NPU
encoder	QNN_DLC	float	Qualcomm® QCS9075	229.033 ms	12 - 42 MB	NPU
encoder	QNN_DLC	float	Snapdragon® 8 Elite For Galaxy Mobile	97.831 ms	0 - 1690 MB	NPU
encoder	QNN_DLC	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	35.091 ms	3 - 1782 MB	NPU
encoder	QNN_DLC	w8a8	Snapdragon® 8 Elite Mobile	40.08 ms	3 - 1711 MB	NPU
encoder	QNN_DLC	w8a8	Snapdragon® X2 Elite	42.762 ms	3 - 3 MB	NPU
encoder	QNN_DLC	w8a8	Snapdragon® X Elite	79.792 ms	3 - 3 MB	NPU
encoder	QNN_DLC	w8a8	Snapdragon® X Elite	79.792 ms	3 - 3 MB	NPU
encoder	QNN_DLC	w8a8	Snapdragon® 8 Gen 3 Mobile	54.566 ms	3 - 2221 MB	NPU
encoder	QNN_DLC	w8a8	Qualcomm® QCS6490	632.146 ms	3 - 12 MB	NPU
encoder	QNN_DLC	w8a8	Qualcomm® QCS8550 (Proxy)	76.587 ms	3 - 6 MB	NPU
encoder	QNN_DLC	w8a8	Snapdragon® 8 Elite For Galaxy Mobile	40.08 ms	3 - 1711 MB	NPU
encoder	QNN_DLC	w8a8	Qualcomm® QCS8450 (Proxy)	136.051 ms	3 - 2227 MB	NPU
encoder	QNN_DLC	w8a8	Snapdragon® 7 Gen 4 Mobile	99.819 ms	3 - 2384 MB	NPU
encoder	QNN_DLC	w8a8	Qualcomm® QCM6690	1251.26 ms	3 - 2713 MB	NPU
encoder	QNN_DLC	w8a8	Qualcomm® SA8295P	92.435 ms	1 - 1759 MB	NPU
encoder	QNN_DLC	w8a8	Qualcomm® QCS9075	93.249 ms	3 - 12 MB	NPU
encoder	QNN_DLC	w8a8	Snapdragon® 7 Gen 4 Mobile	99.819 ms	3 - 2384 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.