MaskRCNN: Optimized for Qualcomm Devices

Mask R-CNN is a machine learning model that extends Faster R-CNN to perform instance segmentation by detecting objects in an image while simultaneously generating a high-quality segmentation mask for each instance. It adds a branch for predicting segmentation masks in parallel with the existing branch for bounding box recognition.

This is based on the implementation of MaskRCNN 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

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

Model Details

Model Type: Model_use_case.semantic_segmentation

Model Stats:

• Model checkpoint: Mask R-CNN ResNet-50 FPN V2
• Input resolution: 800x800
• Number of output classes: 91
• Number of parameters: 46.4M
• Model size (float): 177 MB

Performance Summary

Model	Runtime	Precision	Chipset	Inference Time (ms)	Peak Memory Range (MB)	Primary Compute Unit
proposal_generator	QNN_DLC	float	Snapdragon® X2 Elite	42.855 ms	7 - 7 MB	NPU
proposal_generator	QNN_DLC	float	Snapdragon® X Elite	91.325 ms	7 - 7 MB	NPU
proposal_generator	QNN_DLC	float	Snapdragon® 8 Gen 3 Mobile	70.051 ms	7 - 2327 MB	NPU
proposal_generator	QNN_DLC	float	Snapdragon® 8 Gen 1 Mobile	152.938 ms	8 - 2736 MB	NPU
proposal_generator	QNN_DLC	float	Qualcomm® QCS8275	354.153 ms	2 - 1758 MB	NPU
proposal_generator	QNN_DLC	float	Qualcomm® QCS8550 (Proxy)	95.709 ms	7 - 16 MB	NPU
proposal_generator	QNN_DLC	float	Qualcomm® QCS8450	152.938 ms	8 - 2736 MB	NPU
proposal_generator	QNN_DLC	float	Snapdragon® 8 Elite Mobile	53.463 ms	0 - 1510 MB	NPU
proposal_generator	QNN_DLC	float	Qualcomm® SA8295P	125.945 ms	3 - 1432 MB	NPU
proposal_generator	QNN_DLC	float	Snapdragon® 8 Elite Gen 5 Mobile	41.81 ms	7 - 1789 MB	NPU
proposal_generator	QNN_DLC	float	Qualcomm® SA7255P	354.153 ms	2 - 1758 MB	NPU
proposal_generator	QNN_DLC	float	Qualcomm® QCS9075	118.978 ms	7 - 71 MB	NPU
proposal_generator	QNN_DLC	float	Qualcomm® QCS8750	53.463 ms	0 - 1510 MB	NPU
proposal_generator	QNN_DLC	float	Qualcomm® QCS7181	91.325 ms	7 - 7 MB	NPU
roi_head	QNN_DLC	float	Snapdragon® X2 Elite	98.085 ms	52 - 52 MB	NPU
roi_head	QNN_DLC	float	Snapdragon® X Elite	250.621 ms	52 - 52 MB	NPU
roi_head	QNN_DLC	float	Snapdragon® 8 Gen 3 Mobile	177.887 ms	13 - 860 MB	NPU
roi_head	QNN_DLC	float	Snapdragon® 8 Gen 1 Mobile	312.774 ms	39 - 960 MB	NPU
roi_head	QNN_DLC	float	Qualcomm® QCS8275	573.715 ms	49 - 744 MB	NPU
roi_head	QNN_DLC	float	Qualcomm® QCS8550 (Proxy)	242.438 ms	52 - 54 MB	NPU
roi_head	QNN_DLC	float	Qualcomm® QCS8450	312.774 ms	39 - 960 MB	NPU
roi_head	QNN_DLC	float	Snapdragon® 8 Elite Mobile	129.478 ms	39 - 730 MB	NPU
roi_head	QNN_DLC	float	Qualcomm® SA8295P	325.17 ms	49 - 846 MB	NPU
roi_head	QNN_DLC	float	Snapdragon® 8 Elite Gen 5 Mobile	96.26 ms	15 - 721 MB	NPU
roi_head	QNN_DLC	float	Qualcomm® SA7255P	573.715 ms	49 - 744 MB	NPU
roi_head	QNN_DLC	float	Qualcomm® QCS9075	324.226 ms	52 - 106 MB	NPU
roi_head	QNN_DLC	float	Qualcomm® QCS8750	129.478 ms	39 - 730 MB	NPU
roi_head	QNN_DLC	float	Qualcomm® QCS7181	250.621 ms	52 - 52 MB	NPU

License

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

References

• Mask R-CNN
• 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.