Segment-Anything-Model-3: Optimized for Qualcomm Devices

SAM3 (Segment Anything with Concepts) extends SAM2 with open-vocabulary segmentation, producing bounding boxes and masks for objects matching a natural-language prompt.

This is based on the implementation of Segment-Anything-Model-3 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

Due to licensing restrictions, we cannot distribute pre-exported model assets for this model. 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

See our repository for Segment-Anything-Model-3 on GitHub for usage instructions.

Model Details

Model Type: Model_use_case.semantic_segmentation

Model Stats:

• Input resolution: 1008x1008
• Model checkpoint: sam3
• Model size (SAM3Backbone) (float): 128 MB
• Model size (SAM3Transformer) (float): 23.7 MB
• Number of parameters (SAM3Backbone): 33.5M
• Number of parameters (SAM3Transformer): 6.22M

Performance Summary

Model	Runtime	Precision	Chipset	Inference Time (ms)	Peak Memory Range (MB)	Primary Compute Unit
head	PRECOMPILED_QNN_ONNX	float	Snapdragon® 8 Elite Gen 5 For Galaxy Mobile	229.812 ms	130 - 142 MB	NPU
head	PRECOMPILED_QNN_ONNX	float	Snapdragon® 8 Elite For Galaxy Mobile	272.419 ms	131 - 142 MB	NPU
head	PRECOMPILED_QNN_ONNX	float	Snapdragon® X2 Elite	208.75 ms	136 - 136 MB	NPU
head	PRECOMPILED_QNN_ONNX	float	Snapdragon® X Elite	471.969 ms	663 - 663 MB	NPU
head	PRECOMPILED_QNN_ONNX	float	Snapdragon® 8 Gen 3 Mobile	359.659 ms	156 - 166 MB	NPU
head	PRECOMPILED_QNN_ONNX	float	Qualcomm® Dragonwing™ IQ-8275	472.887 ms	106 - 217 MB	NPU
head	PRECOMPILED_QNN_ONNX	float	Qualcomm® Dragonwing™ IQ-9075	485.223 ms	106 - 216 MB	NPU
head	PRECOMPILED_QNN_ONNX	float	Qualcomm® Dragonwing™ IQ-X7181	471.969 ms	663 - 663 MB	NPU
head	PRECOMPILED_QNN_ONNX	float	Qualcomm® Dragonwing™ Q-8750	272.419 ms	131 - 142 MB	NPU
head	QNN_CONTEXT_BINARY	float	Snapdragon® 8 Elite Gen 5 For Galaxy Mobile	224.248 ms	36 - 44 MB	NPU
head	QNN_CONTEXT_BINARY	float	Snapdragon® 8 Elite For Galaxy Mobile	267.463 ms	86 - 94 MB	NPU
head	QNN_CONTEXT_BINARY	float	Snapdragon® X2 Elite	209.59 ms	107 - 107 MB	NPU
head	QNN_CONTEXT_BINARY	float	Snapdragon® X Elite	471.958 ms	107 - 107 MB	NPU
head	QNN_CONTEXT_BINARY	float	Snapdragon® 8 Gen 3 Mobile	355.987 ms	105 - 112 MB	NPU
head	QNN_CONTEXT_BINARY	float	Qualcomm® Dragonwing™ IQ-8275	472.849 ms	106 - 279 MB	NPU
head	QNN_CONTEXT_BINARY	float	Qualcomm® SA8650P	516.497 ms	97 - 107 MB	NPU
head	QNN_CONTEXT_BINARY	float	Qualcomm® SA8255P	516.497 ms	97 - 107 MB	NPU
head	QNN_CONTEXT_BINARY	float	Qualcomm® Dragonwing™ IQ-9075	485.687 ms	106 - 278 MB	NPU
head	QNN_CONTEXT_BINARY	float	Qualcomm® Dragonwing™ IQ-X7181	471.958 ms	107 - 107 MB	NPU
head	QNN_CONTEXT_BINARY	float	Qualcomm® Dragonwing™ Q-8750	267.463 ms	86 - 94 MB	NPU
head	QNN_CONTEXT_BINARY	float	Qualcomm® SA8295P	582.42 ms	81 - 86 MB	NPU
vision_backbone	PRECOMPILED_QNN_ONNX	float	Snapdragon® 8 Elite Gen 5 For Galaxy Mobile	1201.787 ms	78 - 90 MB	NPU
vision_backbone	PRECOMPILED_QNN_ONNX	float	Snapdragon® 8 Elite For Galaxy Mobile	1293.754 ms	8 - 20 MB	NPU
vision_backbone	PRECOMPILED_QNN_ONNX	float	Snapdragon® X2 Elite	1005.123 ms	265 - 265 MB	NPU
vision_backbone	PRECOMPILED_QNN_ONNX	float	Snapdragon® X Elite	2108.201 ms	951 - 951 MB	NPU
vision_backbone	PRECOMPILED_QNN_ONNX	float	Snapdragon® 8 Gen 3 Mobile	1648.033 ms	20 - 29 MB	NPU
vision_backbone	PRECOMPILED_QNN_ONNX	float	Qualcomm® Dragonwing™ IQ-8275	2461.294 ms	16 - 32 MB	NPU
vision_backbone	PRECOMPILED_QNN_ONNX	float	Qualcomm® Dragonwing™ IQ-9075	2413.725 ms	12 - 27 MB	NPU
vision_backbone	PRECOMPILED_QNN_ONNX	float	Qualcomm® Dragonwing™ IQ-X7181	2108.201 ms	951 - 951 MB	NPU
vision_backbone	PRECOMPILED_QNN_ONNX	float	Qualcomm® Dragonwing™ Q-8750	1293.754 ms	8 - 20 MB	NPU
vision_backbone	QNN_CONTEXT_BINARY	float	Snapdragon® 8 Elite Gen 5 For Galaxy Mobile	1090.968 ms	6 - 19 MB	NPU
vision_backbone	QNN_CONTEXT_BINARY	float	Snapdragon® 8 Elite For Galaxy Mobile	1290.506 ms	0 - 8 MB	NPU
vision_backbone	QNN_CONTEXT_BINARY	float	Snapdragon® X2 Elite	1007.88 ms	12 - 12 MB	NPU
vision_backbone	QNN_CONTEXT_BINARY	float	Snapdragon® X Elite	2150.561 ms	12 - 12 MB	NPU
vision_backbone	QNN_CONTEXT_BINARY	float	Snapdragon® 8 Gen 3 Mobile	1616.635 ms	12 - 24 MB	NPU
vision_backbone	QNN_CONTEXT_BINARY	float	Qualcomm® Dragonwing™ IQ-8275	2459.818 ms	12 - 133 MB	NPU
vision_backbone	QNN_CONTEXT_BINARY	float	Qualcomm® SA8650P	2520.912 ms	1 - 11 MB	NPU
vision_backbone	QNN_CONTEXT_BINARY	float	Qualcomm® SA8255P	2520.912 ms	1 - 11 MB	NPU
vision_backbone	QNN_CONTEXT_BINARY	float	Qualcomm® Dragonwing™ IQ-9075	2413.703 ms	12 - 132 MB	NPU
vision_backbone	QNN_CONTEXT_BINARY	float	Qualcomm® Dragonwing™ IQ-X7181	2150.561 ms	12 - 12 MB	NPU
vision_backbone	QNN_CONTEXT_BINARY	float	Qualcomm® Dragonwing™ Q-8750	1290.506 ms	0 - 8 MB	NPU
vision_backbone	QNN_CONTEXT_BINARY	float	Qualcomm® SA8295P	2799.92 ms	1 - 11 MB	NPU

License

• The license for the original implementation of Segment-Anything-Model-3 can be found here.

References

• SAM 3: Segment Anything with Concepts
• 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.