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 2.43	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® 8 Elite Gen 5 Mobile	54.714 ms	7 - 1325 MB	NPU
proposal_generator	QNN_DLC	float	Snapdragon® X2 Elite	58.092 ms	7 - 7 MB	NPU
proposal_generator	QNN_DLC	float	Snapdragon® X Elite	140.594 ms	7 - 7 MB	NPU
proposal_generator	QNN_DLC	float	Snapdragon® 8 Gen 3 Mobile	107.115 ms	7 - 1391 MB	NPU
proposal_generator	QNN_DLC	float	Qualcomm® QCS8275 (Proxy)	416.068 ms	0 - 1193 MB	NPU
proposal_generator	QNN_DLC	float	Qualcomm® QCS8550 (Proxy)	149.815 ms	7 - 10 MB	NPU
proposal_generator	QNN_DLC	float	Qualcomm® SA8775P	167.146 ms	2 - 1195 MB	NPU
proposal_generator	QNN_DLC	float	Qualcomm® QCS9075	172.276 ms	7 - 71 MB	NPU
proposal_generator	QNN_DLC	float	Qualcomm® QCS8450 (Proxy)	205.605 ms	8 - 1353 MB	NPU
proposal_generator	QNN_DLC	float	Qualcomm® SA7255P	416.068 ms	0 - 1193 MB	NPU
proposal_generator	QNN_DLC	float	Qualcomm® SA8295P	168.632 ms	0 - 1140 MB	NPU
proposal_generator	QNN_DLC	float	Snapdragon® 8 Elite For Galaxy Mobile	71.309 ms	7 - 1303 MB	NPU
roi_head	QNN_DLC	float	Snapdragon® 8 Elite Gen 5 Mobile	94.933 ms	45 - 714 MB	NPU
roi_head	QNN_DLC	float	Snapdragon® X2 Elite	99.942 ms	52 - 52 MB	NPU
roi_head	QNN_DLC	float	Snapdragon® X Elite	239.657 ms	52 - 52 MB	NPU
roi_head	QNN_DLC	float	Snapdragon® 8 Gen 3 Mobile	178.724 ms	16 - 831 MB	NPU
roi_head	QNN_DLC	float	Qualcomm® QCS8275 (Proxy)	582.307 ms	42 - 858 MB	NPU
roi_head	QNN_DLC	float	Qualcomm® QCS8550 (Proxy)	252.096 ms	52 - 54 MB	NPU
roi_head	QNN_DLC	float	Qualcomm® SA8775P	274.199 ms	47 - 842 MB	NPU
roi_head	QNN_DLC	float	Qualcomm® QCS9075	332.746 ms	52 - 106 MB	NPU
roi_head	QNN_DLC	float	Qualcomm® QCS8450 (Proxy)	315.454 ms	39 - 931 MB	NPU
roi_head	QNN_DLC	float	Qualcomm® SA7255P	582.307 ms	42 - 858 MB	NPU
roi_head	QNN_DLC	float	Qualcomm® SA8295P	302.361 ms	49 - 985 MB	NPU
roi_head	QNN_DLC	float	Snapdragon® 8 Elite For Galaxy Mobile	127.0 ms	36 - 824 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.