MobileSam: Optimized for Qualcomm Devices

Transformer based encoder-decoder where prompts specify what to segment in an image thereby allowing segmentation without the need for additional training. The image encoder generates embeddings and the lightweight decoder operates on the embeddings for point and mask based image segmentation.

This is based on the implementation of MobileSam 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 1.25.0	Download
QNN_DLC	float	Universal	QAIRT 2.45	Download
TFLITE	float	Universal	QAIRT 2.45	Download

For more device-specific assets and performance metrics, visit MobileSam 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 MobileSam on GitHub for usage instructions.

Model Details

Model Type: Model_use_case.semantic_segmentation

Model Stats:

• Model checkpoint: vit_t
• Input resolution: 720p (720x1280)
• Number of parameters (sam_encoder): 6.95M
• Model size (sam_encoder) (float): 26.6 MB
• Number of parameters (sam_decoder): 6.16M
• Model size (sam_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® X2 Elite	2.635 ms	209 - 209 MB	NPU
decoder	ONNX	float	Snapdragon® X Elite	6.168 ms	147 - 147 MB	NPU
decoder	ONNX	float	Snapdragon® 8 Gen 3 Mobile	4.15 ms	2 - 234 MB	NPU
decoder	ONNX	float	Snapdragon® 8 Gen 1 Mobile	14.299 ms	5 - 225 MB	NPU
decoder	ONNX	float	Qualcomm® QCS8550 (Proxy)	5.917 ms	0 - 19 MB	NPU
decoder	ONNX	float	Qualcomm® QCS8450	14.299 ms	5 - 225 MB	NPU
decoder	ONNX	float	Snapdragon® 8 Elite Mobile	3.173 ms	0 - 234 MB	NPU
decoder	ONNX	float	Snapdragon® 8 Elite Gen 5 Mobile	2.531 ms	3 - 210 MB	NPU
decoder	ONNX	float	Qualcomm® QCS9075	7.129 ms	4 - 49 MB	NPU
decoder	ONNX	float	Qualcomm® QCS8750	3.173 ms	0 - 234 MB	NPU
decoder	ONNX	float	Qualcomm® QCS7181	6.168 ms	147 - 147 MB	NPU
decoder	QNN_DLC	float	Snapdragon® X2 Elite	2.97 ms	4 - 4 MB	NPU
decoder	QNN_DLC	float	Snapdragon® X Elite	5.722 ms	4 - 4 MB	NPU
decoder	QNN_DLC	float	Snapdragon® 8 Gen 3 Mobile	3.732 ms	4 - 223 MB	NPU
decoder	QNN_DLC	float	Snapdragon® 8 Gen 1 Mobile	11.626 ms	0 - 251 MB	NPU
decoder	QNN_DLC	float	Qualcomm® QCS8275	12.614 ms	0 - 192 MB	NPU
decoder	QNN_DLC	float	Qualcomm® QCS8550 (Proxy)	5.307 ms	4 - 132 MB	NPU
decoder	QNN_DLC	float	Qualcomm® QCS8450	11.626 ms	0 - 251 MB	NPU
decoder	QNN_DLC	float	Snapdragon® 8 Elite Mobile	2.833 ms	0 - 225 MB	NPU
decoder	QNN_DLC	float	Qualcomm® SA8295P	7.714 ms	0 - 190 MB	NPU
decoder	QNN_DLC	float	Snapdragon® 8 Elite Gen 5 Mobile	2.526 ms	4 - 196 MB	NPU
decoder	QNN_DLC	float	Qualcomm® SA7255P	12.614 ms	0 - 192 MB	NPU
decoder	QNN_DLC	float	Qualcomm® QCS9075	6.49 ms	4 - 10 MB	NPU
decoder	QNN_DLC	float	Qualcomm® QCS8750	2.833 ms	0 - 225 MB	NPU
decoder	QNN_DLC	float	Qualcomm® QCS7181	5.722 ms	4 - 4 MB	NPU
decoder	TFLITE	float	Snapdragon® 8 Gen 3 Mobile	3.42 ms	0 - 214 MB	NPU
decoder	TFLITE	float	Snapdragon® 8 Gen 1 Mobile	7.991 ms	0 - 217 MB	NPU
decoder	TFLITE	float	Qualcomm® QCS8275	11.888 ms	0 - 186 MB	NPU
decoder	TFLITE	float	Qualcomm® QCS8550 (Proxy)	4.937 ms	0 - 8 MB	NPU
decoder	TFLITE	float	Qualcomm® SA8775P	38.522 ms	1 - 34 MB	GPU
decoder	TFLITE	float	Qualcomm® SA8650P	38.522 ms	1 - 34 MB	GPU
decoder	TFLITE	float	Qualcomm® SA8255P	38.522 ms	1 - 34 MB	GPU
decoder	TFLITE	float	Qualcomm® QCS8450	7.991 ms	0 - 217 MB	NPU
decoder	TFLITE	float	Snapdragon® 8 Elite Mobile	2.591 ms	0 - 183 MB	NPU
decoder	TFLITE	float	Qualcomm® SA8295P	7.261 ms	0 - 183 MB	NPU
decoder	TFLITE	float	Snapdragon® 8 Elite Gen 5 Mobile	2.159 ms	0 - 187 MB	NPU
decoder	TFLITE	float	Qualcomm® SA7255P	11.888 ms	0 - 186 MB	NPU
decoder	TFLITE	float	Qualcomm® QCS9075	5.969 ms	0 - 19 MB	NPU
decoder	TFLITE	float	Qualcomm® QCS8750	2.591 ms	0 - 183 MB	NPU
encoder	ONNX	float	Snapdragon® X2 Elite	91.844 ms	201 - 201 MB	NPU
encoder	ONNX	float	Snapdragon® X Elite	170.625 ms	170 - 170 MB	NPU
encoder	ONNX	float	Snapdragon® 8 Gen 3 Mobile	125.122 ms	21 - 1099 MB	NPU
encoder	ONNX	float	Snapdragon® 8 Gen 1 Mobile	407.007 ms	22 - 1118 MB	NPU
encoder	ONNX	float	Qualcomm® QCS8550 (Proxy)	172.174 ms	12 - 45 MB	NPU
encoder	ONNX	float	Qualcomm® QCS8450	407.007 ms	22 - 1118 MB	NPU
encoder	ONNX	float	Snapdragon® 8 Elite Mobile	99.607 ms	9 - 922 MB	NPU
encoder	ONNX	float	Snapdragon® 8 Elite Gen 5 Mobile	81.662 ms	8 - 1032 MB	NPU
encoder	ONNX	float	Qualcomm® QCS9075	195.438 ms	12 - 57 MB	NPU
encoder	ONNX	float	Qualcomm® QCS8750	99.607 ms	9 - 922 MB	NPU
encoder	ONNX	float	Qualcomm® QCS7181	170.625 ms	170 - 170 MB	NPU
encoder	QNN_DLC	float	Snapdragon® X2 Elite	45.856 ms	12 - 12 MB	NPU
encoder	QNN_DLC	float	Snapdragon® X Elite	103.032 ms	12 - 12 MB	NPU
encoder	QNN_DLC	float	Snapdragon® 8 Gen 3 Mobile	65.901 ms	12 - 1329 MB	NPU
encoder	QNN_DLC	float	Snapdragon® 8 Gen 1 Mobile	329.08 ms	12 - 1342 MB	NPU
encoder	QNN_DLC	float	Qualcomm® QCS8275	207.642 ms	1 - 1126 MB	NPU
encoder	QNN_DLC	float	Qualcomm® QCS8550 (Proxy)	99.176 ms	12 - 891 MB	NPU
encoder	QNN_DLC	float	Qualcomm® QCS8450	329.08 ms	12 - 1342 MB	NPU
encoder	QNN_DLC	float	Snapdragon® 8 Elite Mobile	48.69 ms	12 - 1129 MB	NPU
encoder	QNN_DLC	float	Qualcomm® SA8295P	324.852 ms	0 - 1556 MB	NPU
encoder	QNN_DLC	float	Snapdragon® 8 Elite Gen 5 Mobile	40.129 ms	12 - 1260 MB	NPU
encoder	QNN_DLC	float	Qualcomm® SA7255P	207.642 ms	1 - 1126 MB	NPU
encoder	QNN_DLC	float	Qualcomm® QCS9075	120.784 ms	14 - 32 MB	NPU
encoder	QNN_DLC	float	Qualcomm® QCS8750	48.69 ms	12 - 1129 MB	NPU
encoder	QNN_DLC	float	Qualcomm® QCS7181	103.032 ms	12 - 12 MB	NPU
encoder	TFLITE	float	Snapdragon® 8 Gen 3 Mobile	165.516 ms	33 - 2325 MB	NPU
encoder	TFLITE	float	Snapdragon® 8 Gen 1 Mobile	512.784 ms	0 - 1933 MB	NPU
encoder	TFLITE	float	Qualcomm® QCS8275	360.203 ms	33 - 1120 MB	NPU
encoder	TFLITE	float	Qualcomm® QCS8550 (Proxy)	222.998 ms	31 - 34 MB	NPU
encoder	TFLITE	float	Qualcomm® SA8775P	740.89 ms	0 - 25 MB	CPU
encoder	TFLITE	float	Qualcomm® SA8650P	740.89 ms	0 - 25 MB	CPU
encoder	TFLITE	float	Qualcomm® SA8255P	740.89 ms	0 - 25 MB	CPU
encoder	TFLITE	float	Qualcomm® QCS8450	512.784 ms	0 - 1933 MB	NPU
encoder	TFLITE	float	Snapdragon® 8 Elite Mobile	128.85 ms	32 - 1106 MB	NPU
encoder	TFLITE	float	Qualcomm® SA8295P	372.611 ms	34 - 1161 MB	NPU
encoder	TFLITE	float	Snapdragon® 8 Elite Gen 5 Mobile	98.436 ms	32 - 1097 MB	NPU
encoder	TFLITE	float	Qualcomm® SA7255P	360.203 ms	33 - 1120 MB	NPU
encoder	TFLITE	float	Qualcomm® QCS9075	291.604 ms	32 - 109 MB	NPU
encoder	TFLITE	float	Qualcomm® QCS8750	128.85 ms	32 - 1106 MB	NPU

License

• The license for the original implementation of MobileSam can be found here.

References

• 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.