Masked-attention Mask Transformer for Universal Image Segmentation
Paper • 2112.01527 • Published • 1
Mask2Former: Optimized for Qualcomm Devices
Mask2Former is a machine learning model that predicts masks and classes of objects in an image.
This is based on the implementation of Mask2Former 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 |
|---|---|---|---|---|
| PRECOMPILED_QNN_ONNX | float | Snapdragon® X2 Elite | QAIRT 2.45, ONNX Runtime 1.27.1 | Download |
| PRECOMPILED_QNN_ONNX | float | Snapdragon® X Elite | QAIRT 2.45, ONNX Runtime 1.27.1 | Download |
| PRECOMPILED_QNN_ONNX | float | Snapdragon® 8 Gen 3 Mobile | QAIRT 2.45, ONNX Runtime 1.27.1 | Download |
| PRECOMPILED_QNN_ONNX | float | Snapdragon® 8 Gen 1 Mobile | QAIRT 2.45, ONNX Runtime 1.27.1 | Download |
| PRECOMPILED_QNN_ONNX | float | Qualcomm® Dragonwing™ IQ-8275 | QAIRT 2.45, ONNX Runtime 1.27.1 | Download |
| PRECOMPILED_QNN_ONNX | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | QAIRT 2.45, ONNX Runtime 1.27.1 | Download |
| PRECOMPILED_QNN_ONNX | float | Qualcomm® Dragonwing™ IQ-9075 | QAIRT 2.45, ONNX Runtime 1.27.1 | Download |
| PRECOMPILED_QNN_ONNX | float | Snapdragon® 8 Elite Mobile | QAIRT 2.45, ONNX Runtime 1.27.1 | Download |
| PRECOMPILED_QNN_ONNX | float | Snapdragon® 8 Elite Gen 5 Mobile | QAIRT 2.45, ONNX Runtime 1.27.1 | Download |
| QNN_CONTEXT_BINARY | float | Snapdragon® X2 Elite | QAIRT 2.45 | Download |
| QNN_CONTEXT_BINARY | float | Snapdragon® X Elite | QAIRT 2.45 | Download |
| QNN_CONTEXT_BINARY | float | Snapdragon® 8 Gen 3 Mobile | QAIRT 2.45 | Download |
| QNN_CONTEXT_BINARY | float | Snapdragon® 8 Gen 1 Mobile | QAIRT 2.45 | Download |
| QNN_CONTEXT_BINARY | float | Qualcomm® Dragonwing™ IQ-8275 | QAIRT 2.45 | Download |
| QNN_CONTEXT_BINARY | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | QAIRT 2.45 | Download |
| QNN_CONTEXT_BINARY | float | Qualcomm® SA8775P | QAIRT 2.45 | Download |
| QNN_CONTEXT_BINARY | float | Qualcomm® Dragonwing™ IQ-9075 | QAIRT 2.45 | Download |
| QNN_CONTEXT_BINARY | float | Qualcomm® SA7255P | QAIRT 2.45 | Download |
| QNN_CONTEXT_BINARY | float | Qualcomm® SA8295P | QAIRT 2.45 | Download |
| QNN_CONTEXT_BINARY | float | Snapdragon® 8 Elite Mobile | QAIRT 2.45 | Download |
| QNN_CONTEXT_BINARY | float | Snapdragon® 8 Elite Gen 5 Mobile | QAIRT 2.45 | Download |
For more device-specific assets and performance metrics, visit Mask2Former 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 Mask2Former on GitHub for usage instructions.
Model Details
Model Type: Model_use_case.semantic_segmentation
Model Stats:
- Input resolution: 384x384
- Model checkpoint: facebook/mask2former-swin-tiny-coco-panoptic
- Number of output classes: 100
Performance Summary
| Model | Runtime | Precision | Chipset | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit |
|---|---|---|---|---|---|---|
| Mask2Former | PRECOMPILED_QNN_ONNX | float | Snapdragon® X2 Elite | 67.245 ms | 10 - 10 MB | NPU |
| Mask2Former | PRECOMPILED_QNN_ONNX | float | Snapdragon® X Elite | 141.868 ms | 109 - 109 MB | NPU |
| Mask2Former | PRECOMPILED_QNN_ONNX | float | Snapdragon® 8 Gen 3 Mobile | 103.795 ms | 9 - 21 MB | NPU |
| Mask2Former | PRECOMPILED_QNN_ONNX | float | Snapdragon® 8 Gen 1 Mobile | 233.174 ms | 9 - 23 MB | NPU |
| Mask2Former | PRECOMPILED_QNN_ONNX | float | Qualcomm® Dragonwing™ IQ-8275 | 139.372 ms | 7 - 13 MB | NPU |
| Mask2Former | PRECOMPILED_QNN_ONNX | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 141.13 ms | 0 - 113 MB | NPU |
| Mask2Former | PRECOMPILED_QNN_ONNX | float | Qualcomm® QCS8450 | 233.174 ms | 9 - 23 MB | NPU |
| Mask2Former | PRECOMPILED_QNN_ONNX | float | Qualcomm® Dragonwing™ IQ-9075 | 146.242 ms | 7 - 12 MB | NPU |
| Mask2Former | PRECOMPILED_QNN_ONNX | float | Qualcomm® Dragonwing™ IQ-X7181 | 141.868 ms | 109 - 109 MB | NPU |
| Mask2Former | PRECOMPILED_QNN_ONNX | float | Qualcomm® Dragonwing™ Q-8750 | 79.772 ms | 10 - 21 MB | NPU |
| Mask2Former | PRECOMPILED_QNN_ONNX | float | Snapdragon® 8 Elite Mobile | 79.772 ms | 10 - 21 MB | NPU |
| Mask2Former | PRECOMPILED_QNN_ONNX | float | Snapdragon® 8 Elite Gen 5 Mobile | 70.127 ms | 9 - 21 MB | NPU |
| Mask2Former | QNN_CONTEXT_BINARY | float | Snapdragon® X2 Elite | 67.443 ms | 2 - 2 MB | NPU |
| Mask2Former | QNN_CONTEXT_BINARY | float | Snapdragon® X Elite | 142.575 ms | 2 - 2 MB | NPU |
| Mask2Former | QNN_CONTEXT_BINARY | float | Snapdragon® 8 Gen 3 Mobile | 104.022 ms | 4 - 12 MB | NPU |
| Mask2Former | QNN_CONTEXT_BINARY | float | Snapdragon® 8 Gen 1 Mobile | 233.262 ms | 1 - 11 MB | NPU |
| Mask2Former | QNN_CONTEXT_BINARY | float | Qualcomm® Dragonwing™ IQ-8275 | 139.254 ms | 2 - 9 MB | NPU |
| Mask2Former | QNN_CONTEXT_BINARY | float | Qualcomm® Dragonwing™ IQ-8275 | 289.917 ms | 2 - 11 MB | NPU |
| Mask2Former | QNN_CONTEXT_BINARY | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 140.937 ms | 2 - 8 MB | NPU |
| Mask2Former | QNN_CONTEXT_BINARY | float | Qualcomm® SA8775P | 146.959 ms | 2 - 11 MB | NPU |
| Mask2Former | QNN_CONTEXT_BINARY | float | Qualcomm® SA8650P | 146.959 ms | 2 - 11 MB | NPU |
| Mask2Former | QNN_CONTEXT_BINARY | float | Qualcomm® SA8255P | 146.959 ms | 2 - 11 MB | NPU |
| Mask2Former | QNN_CONTEXT_BINARY | float | Qualcomm® QCS8450 | 233.262 ms | 1 - 11 MB | NPU |
| Mask2Former | QNN_CONTEXT_BINARY | float | Qualcomm® Dragonwing™ IQ-9075 | 183.663 ms | 2 - 9 MB | NPU |
| Mask2Former | QNN_CONTEXT_BINARY | float | Qualcomm® Dragonwing™ IQ-X7181 | 142.575 ms | 2 - 2 MB | NPU |
| Mask2Former | QNN_CONTEXT_BINARY | float | Qualcomm® Dragonwing™ Q-8750 | 80.348 ms | 2 - 12 MB | NPU |
| Mask2Former | QNN_CONTEXT_BINARY | float | Qualcomm® SA7255P | 289.917 ms | 2 - 11 MB | NPU |
| Mask2Former | QNN_CONTEXT_BINARY | float | Qualcomm® SA8295P | 181.933 ms | 2 - 7 MB | NPU |
| Mask2Former | QNN_CONTEXT_BINARY | float | Snapdragon® 8 Elite Mobile | 80.348 ms | 2 - 12 MB | NPU |
| Mask2Former | QNN_CONTEXT_BINARY | float | Snapdragon® 8 Elite Gen 5 Mobile | 70.01 ms | 2 - 12 MB | NPU |
License
- The license for the original implementation of Mask2Former can be found here.
References
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.