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 2.42, ONNX Runtime 1.24.3	Download
QNN_DLC	float	Universal	QAIRT 2.43	Download
TFLITE	float	Universal	QAIRT 2.43, TFLite 2.19.1	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® 8 Elite Gen 5 Mobile	2.948 ms	0 - 243 MB	NPU
decoder	ONNX	float	Snapdragon® X2 Elite	3.023 ms	11 - 11 MB	NPU
decoder	ONNX	float	Snapdragon® X Elite	7.621 ms	11 - 11 MB	NPU
decoder	ONNX	float	Snapdragon® 8 Gen 3 Mobile	5.008 ms	5 - 350 MB	NPU
decoder	ONNX	float	Qualcomm® QCS8550 (Proxy)	7.11 ms	0 - 15 MB	NPU
decoder	ONNX	float	Qualcomm® QCS9075	8.436 ms	4 - 7 MB	NPU
decoder	ONNX	float	Snapdragon® 8 Elite For Galaxy Mobile	3.655 ms	0 - 222 MB	NPU
decoder	QNN_DLC	float	Snapdragon® 8 Elite Gen 5 Mobile	2.379 ms	3 - 203 MB	NPU
decoder	QNN_DLC	float	Snapdragon® X2 Elite	2.849 ms	4 - 4 MB	NPU
decoder	QNN_DLC	float	Snapdragon® X Elite	5.973 ms	4 - 4 MB	NPU
decoder	QNN_DLC	float	Snapdragon® 8 Gen 3 Mobile	3.844 ms	0 - 311 MB	NPU
decoder	QNN_DLC	float	Qualcomm® QCS8275 (Proxy)	12.991 ms	1 - 283 MB	NPU
decoder	QNN_DLC	float	Qualcomm® QCS8550 (Proxy)	5.527 ms	4 - 6 MB	NPU
decoder	QNN_DLC	float	Qualcomm® SA8775P	6.478 ms	0 - 284 MB	NPU
decoder	QNN_DLC	float	Qualcomm® QCS9075	6.913 ms	4 - 10 MB	NPU
decoder	QNN_DLC	float	Qualcomm® QCS8450 (Proxy)	9.63 ms	2 - 227 MB	NPU
decoder	QNN_DLC	float	Qualcomm® SA7255P	12.991 ms	1 - 283 MB	NPU
decoder	QNN_DLC	float	Qualcomm® SA8295P	8.029 ms	0 - 196 MB	NPU
decoder	QNN_DLC	float	Snapdragon® 8 Elite For Galaxy Mobile	2.801 ms	0 - 267 MB	NPU
decoder	TFLITE	float	Snapdragon® 8 Elite Gen 5 Mobile	2.387 ms	0 - 206 MB	NPU
decoder	TFLITE	float	Snapdragon® 8 Gen 3 Mobile	3.904 ms	0 - 237 MB	NPU
decoder	TFLITE	float	Qualcomm® QCS8275 (Proxy)	13.052 ms	0 - 201 MB	NPU
decoder	TFLITE	float	Qualcomm® QCS8550 (Proxy)	5.649 ms	0 - 8 MB	NPU
decoder	TFLITE	float	Qualcomm® SA8775P	6.692 ms	0 - 203 MB	NPU
decoder	TFLITE	float	Qualcomm® QCS9075	6.943 ms	0 - 19 MB	NPU
decoder	TFLITE	float	Qualcomm® QCS8450 (Proxy)	9.518 ms	0 - 224 MB	NPU
decoder	TFLITE	float	Qualcomm® SA7255P	13.052 ms	0 - 201 MB	NPU
decoder	TFLITE	float	Qualcomm® SA8295P	8.088 ms	0 - 200 MB	NPU
decoder	TFLITE	float	Snapdragon® 8 Elite For Galaxy Mobile	2.939 ms	0 - 205 MB	NPU
encoder	ONNX	float	Snapdragon® 8 Elite Gen 5 Mobile	84.094 ms	20 - 1522 MB	NPU
encoder	ONNX	float	Snapdragon® X2 Elite	96.11 ms	14 - 14 MB	NPU
encoder	ONNX	float	Snapdragon® X Elite	172.164 ms	12 - 12 MB	NPU
encoder	ONNX	float	Snapdragon® 8 Gen 3 Mobile	125.916 ms	20 - 3043 MB	NPU
encoder	ONNX	float	Qualcomm® QCS8550 (Proxy)	177.901 ms	12 - 40 MB	NPU
encoder	ONNX	float	Qualcomm® QCS9075	200.498 ms	12 - 15 MB	NPU
encoder	ONNX	float	Snapdragon® 8 Elite For Galaxy Mobile	100.485 ms	9 - 1390 MB	NPU
encoder	QNN_DLC	float	Snapdragon® 8 Elite Gen 5 Mobile	48.926 ms	12 - 1324 MB	NPU
encoder	QNN_DLC	float	Snapdragon® X2 Elite	51.713 ms	12 - 12 MB	NPU
encoder	QNN_DLC	float	Snapdragon® X Elite	100.189 ms	12 - 12 MB	NPU
encoder	QNN_DLC	float	Snapdragon® 8 Gen 3 Mobile	66.047 ms	10 - 1628 MB	NPU
encoder	QNN_DLC	float	Qualcomm® QCS8275 (Proxy)	209.31 ms	0 - 1210 MB	NPU
encoder	QNN_DLC	float	Qualcomm® QCS8550 (Proxy)	97.222 ms	12 - 15 MB	NPU
encoder	QNN_DLC	float	Qualcomm® SA8775P	103.074 ms	1 - 1209 MB	NPU
encoder	QNN_DLC	float	Qualcomm® QCS9075	121.864 ms	12 - 30 MB	NPU
encoder	QNN_DLC	float	Qualcomm® QCS8450 (Proxy)	469.913 ms	11 - 1740 MB	NPU
encoder	QNN_DLC	float	Qualcomm® SA7255P	209.31 ms	0 - 1210 MB	NPU
encoder	QNN_DLC	float	Qualcomm® SA8295P	474.209 ms	0 - 1258 MB	NPU
encoder	QNN_DLC	float	Snapdragon® 8 Elite For Galaxy Mobile	49.637 ms	12 - 1215 MB	NPU
encoder	TFLITE	float	Snapdragon® 8 Elite Gen 5 Mobile	104.259 ms	32 - 1233 MB	NPU
encoder	TFLITE	float	Snapdragon® 8 Gen 3 Mobile	167.58 ms	17 - 1923 MB	NPU
encoder	TFLITE	float	Qualcomm® QCS8275 (Proxy)	360.435 ms	29 - 1180 MB	NPU
encoder	TFLITE	float	Qualcomm® QCS8550 (Proxy)	216.954 ms	33 - 37 MB	NPU
encoder	TFLITE	float	Qualcomm® SA8775P	240.406 ms	33 - 1227 MB	NPU
encoder	TFLITE	float	Qualcomm® QCS9075	289.564 ms	32 - 109 MB	NPU
encoder	TFLITE	float	Qualcomm® QCS8450 (Proxy)	670.636 ms	23 - 2068 MB	NPU
encoder	TFLITE	float	Qualcomm® SA7255P	360.435 ms	29 - 1180 MB	NPU
encoder	TFLITE	float	Qualcomm® SA8295P	523.394 ms	19 - 1181 MB	NPU
encoder	TFLITE	float	Snapdragon® 8 Elite For Galaxy Mobile	131.152 ms	32 - 1235 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.