Scaling Open-Vocabulary Object Detection
Paper • 2306.09683 • Published • 16
OWL-V2: Optimized for Qualcomm Devices
OWL-V2 (Open-World Localization with Vision Transformers) is an open-vocabulary object detector that uses a CLIP-based ViT-B/16 backbone. Given an image and one or more free-form text queries, the model predicts bounding boxes and confidence scores for each query.
This is based on the implementation of OWL-V2 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® 8 Elite Gen 5 For Galaxy Mobile | QAIRT 2.50, ONNX Runtime 1.30.0 | Download |
| PRECOMPILED_QNN_ONNX | float | Snapdragon® 8 Elite For Galaxy Mobile | QAIRT 2.50, ONNX Runtime 1.30.0 | Download |
| PRECOMPILED_QNN_ONNX | float | Snapdragon® X2 Elite | QAIRT 2.50, ONNX Runtime 1.30.0 | Download |
| PRECOMPILED_QNN_ONNX | float | Snapdragon® X Elite | QAIRT 2.50, ONNX Runtime 1.30.0 | Download |
| PRECOMPILED_QNN_ONNX | float | Snapdragon® 8 Gen 3 Mobile | QAIRT 2.50, ONNX Runtime 1.30.0 | Download |
| PRECOMPILED_QNN_ONNX | float | Snapdragon® 8 Gen 1 Mobile | QAIRT 2.50, ONNX Runtime 1.30.0 | Download |
| PRECOMPILED_QNN_ONNX | float | Qualcomm® Dragonwing™ IQ-8275 | QAIRT 2.50, ONNX Runtime 1.30.0 | Download |
| PRECOMPILED_QNN_ONNX | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | QAIRT 2.50, ONNX Runtime 1.30.0 | Download |
| PRECOMPILED_QNN_ONNX | float | Qualcomm® Dragonwing™ IQ-9075 | QAIRT 2.50, ONNX Runtime 1.30.0 | Download |
| QNN_CONTEXT_BINARY | float | Snapdragon® 8 Elite Gen 5 For Galaxy Mobile | QAIRT 2.50 | Download |
| QNN_CONTEXT_BINARY | float | Snapdragon® 8 Elite For Galaxy Mobile | QAIRT 2.50 | Download |
| QNN_CONTEXT_BINARY | float | Snapdragon® X2 Elite | QAIRT 2.50 | Download |
| QNN_CONTEXT_BINARY | float | Snapdragon® X Elite | QAIRT 2.50 | Download |
| QNN_CONTEXT_BINARY | float | Snapdragon® 8 Gen 3 Mobile | QAIRT 2.50 | Download |
| QNN_CONTEXT_BINARY | float | Snapdragon® 8 Gen 1 Mobile | QAIRT 2.50 | Download |
| QNN_CONTEXT_BINARY | float | Qualcomm® Dragonwing™ IQ-8275 | QAIRT 2.50 | Download |
| QNN_CONTEXT_BINARY | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | QAIRT 2.50 | Download |
| QNN_CONTEXT_BINARY | float | Qualcomm® SA8775P | QAIRT 2.50 | Download |
| QNN_CONTEXT_BINARY | float | Qualcomm® Dragonwing™ IQ-9075 | QAIRT 2.50 | Download |
| QNN_CONTEXT_BINARY | float | Qualcomm® SA7255P | QAIRT 2.50 | Download |
| QNN_CONTEXT_BINARY | float | Qualcomm® SA8295P | QAIRT 2.50 | Download |
For more device-specific assets and performance metrics, visit OWL-V2 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 OWL-V2 on GitHub for usage instructions.
Model Details
Model Type: Model_use_case.object_detection
Model Stats:
- Input image resolution: 960x960
- Model checkpoint: google/owlv2-base-patch16-ensemble
- Model size (float): 620 MB
- Number of parameters: 155M
Performance Summary
| Model | Runtime | Precision | Chipset | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit |
|---|---|---|---|---|---|---|
| text | PRECOMPILED_QNN_ONNX | float | Snapdragon® 8 Elite Gen 5 For Galaxy Mobile | 2.791 ms | 3 - 14 MB | NPU |
| text | PRECOMPILED_QNN_ONNX | float | Snapdragon® 8 Elite For Galaxy Mobile | 3.357 ms | 4 - 15 MB | NPU |
| text | PRECOMPILED_QNN_ONNX | float | Snapdragon® X2 Elite | 2.612 ms | 11 - 11 MB | NPU |
| text | PRECOMPILED_QNN_ONNX | float | Snapdragon® X Elite | 5.332 ms | 116 - 116 MB | NPU |
| text | PRECOMPILED_QNN_ONNX | float | Snapdragon® 8 Gen 3 Mobile | 3.939 ms | 11 - 21 MB | NPU |
| text | PRECOMPILED_QNN_ONNX | float | Snapdragon® 8 Gen 1 Mobile | 10.509 ms | 11 - 25 MB | NPU |
| text | PRECOMPILED_QNN_ONNX | float | Qualcomm® Dragonwing™ IQ-8275 | 7.413 ms | 10 - 25 MB | NPU |
| text | PRECOMPILED_QNN_ONNX | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 5.344 ms | 0 - 118 MB | NPU |
| text | PRECOMPILED_QNN_ONNX | float | Qualcomm® QCS8450 | 10.509 ms | 11 - 25 MB | NPU |
| text | PRECOMPILED_QNN_ONNX | float | Qualcomm® Dragonwing™ IQ-9075 | 6.798 ms | 11 - 24 MB | NPU |
| text | PRECOMPILED_QNN_ONNX | float | Qualcomm® Dragonwing™ IQ-X7181 | 5.332 ms | 116 - 116 MB | NPU |
| text | PRECOMPILED_QNN_ONNX | float | Qualcomm® Dragonwing™ Q-8750 | 3.357 ms | 4 - 15 MB | NPU |
| text | QNN_CONTEXT_BINARY | float | Snapdragon® 8 Elite Gen 5 For Galaxy Mobile | 2.789 ms | 10 - 17 MB | NPU |
| text | QNN_CONTEXT_BINARY | float | Snapdragon® 8 Elite For Galaxy Mobile | 3.303 ms | 0 - 11 MB | NPU |
| text | QNN_CONTEXT_BINARY | float | Snapdragon® X2 Elite | 3.076 ms | 11 - 11 MB | NPU |
| text | QNN_CONTEXT_BINARY | float | Snapdragon® X Elite | 5.536 ms | 11 - 11 MB | NPU |
| text | QNN_CONTEXT_BINARY | float | Snapdragon® 8 Gen 3 Mobile | 3.911 ms | 11 - 18 MB | NPU |
| text | QNN_CONTEXT_BINARY | float | Snapdragon® 8 Gen 1 Mobile | 10.492 ms | 10 - 20 MB | NPU |
| text | QNN_CONTEXT_BINARY | float | Qualcomm® Dragonwing™ IQ-8275 | 7.356 ms | 11 - 24 MB | NPU |
| text | QNN_CONTEXT_BINARY | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 5.268 ms | 11 - 12 MB | NPU |
| text | QNN_CONTEXT_BINARY | float | Qualcomm® SA8650P | 7.288 ms | 1 - 10 MB | NPU |
| text | QNN_CONTEXT_BINARY | float | Qualcomm® SA8255P | 7.288 ms | 1 - 10 MB | NPU |
| text | QNN_CONTEXT_BINARY | float | Qualcomm® QCS8450 | 10.492 ms | 10 - 20 MB | NPU |
| text | QNN_CONTEXT_BINARY | float | Qualcomm® Dragonwing™ IQ-9075 | 6.741 ms | 13 - 25 MB | NPU |
| text | QNN_CONTEXT_BINARY | float | Qualcomm® Dragonwing™ IQ-X7181 | 5.536 ms | 11 - 11 MB | NPU |
| text | QNN_CONTEXT_BINARY | float | Qualcomm® Dragonwing™ Q-8750 | 3.303 ms | 0 - 11 MB | NPU |
| text | QNN_CONTEXT_BINARY | float | Qualcomm® SA8295P | 11.963 ms | 0 - 6 MB | NPU |
| vision | PRECOMPILED_QNN_ONNX | float | Snapdragon® 8 Elite Gen 5 For Galaxy Mobile | 444.429 ms | 25 - 38 MB | NPU |
| vision | PRECOMPILED_QNN_ONNX | float | Snapdragon® 8 Elite For Galaxy Mobile | 543.73 ms | 24 - 36 MB | NPU |
| vision | PRECOMPILED_QNN_ONNX | float | Snapdragon® X2 Elite | 427.002 ms | 33 - 33 MB | NPU |
| vision | PRECOMPILED_QNN_ONNX | float | Snapdragon® X Elite | 973.46 ms | 295 - 295 MB | NPU |
| vision | PRECOMPILED_QNN_ONNX | float | Snapdragon® 8 Gen 3 Mobile | 751.321 ms | 24 - 37 MB | NPU |
| vision | PRECOMPILED_QNN_ONNX | float | Snapdragon® 8 Gen 1 Mobile | 1450.561 ms | 11 - 26 MB | NPU |
| vision | PRECOMPILED_QNN_ONNX | float | Qualcomm® Dragonwing™ IQ-8275 | 939.17 ms | 10 - 25 MB | NPU |
| vision | PRECOMPILED_QNN_ONNX | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 1055.295 ms | 0 - 310 MB | NPU |
| vision | PRECOMPILED_QNN_ONNX | float | Qualcomm® QCS8450 | 1450.561 ms | 11 - 26 MB | NPU |
| vision | PRECOMPILED_QNN_ONNX | float | Qualcomm® Dragonwing™ IQ-9075 | 989.445 ms | 10 - 24 MB | NPU |
| vision | PRECOMPILED_QNN_ONNX | float | Qualcomm® Dragonwing™ IQ-X7181 | 973.46 ms | 295 - 295 MB | NPU |
| vision | PRECOMPILED_QNN_ONNX | float | Qualcomm® Dragonwing™ Q-8750 | 543.73 ms | 24 - 36 MB | NPU |
| vision | QNN_CONTEXT_BINARY | float | Snapdragon® 8 Elite Gen 5 For Galaxy Mobile | 431.661 ms | 11 - 19 MB | NPU |
| vision | QNN_CONTEXT_BINARY | float | Snapdragon® 8 Elite For Galaxy Mobile | 527.153 ms | 11 - 22 MB | NPU |
| vision | QNN_CONTEXT_BINARY | float | Snapdragon® X2 Elite | 430.159 ms | 11 - 11 MB | NPU |
| vision | QNN_CONTEXT_BINARY | float | Snapdragon® X Elite | 958.064 ms | 11 - 11 MB | NPU |
| vision | QNN_CONTEXT_BINARY | float | Snapdragon® 8 Gen 3 Mobile | 748.393 ms | 13 - 20 MB | NPU |
| vision | QNN_CONTEXT_BINARY | float | Snapdragon® 8 Gen 1 Mobile | 1400.762 ms | 11 - 25 MB | NPU |
| vision | QNN_CONTEXT_BINARY | float | Qualcomm® Dragonwing™ IQ-8275 | 938.599 ms | 11 - 35 MB | NPU |
| vision | QNN_CONTEXT_BINARY | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 1049.244 ms | 11 - 12 MB | NPU |
| vision | QNN_CONTEXT_BINARY | float | Qualcomm® SA8650P | 1268.382 ms | 3 - 11 MB | NPU |
| vision | QNN_CONTEXT_BINARY | float | Qualcomm® SA8255P | 1268.382 ms | 3 - 11 MB | NPU |
| vision | QNN_CONTEXT_BINARY | float | Qualcomm® QCS8450 | 1400.762 ms | 11 - 25 MB | NPU |
| vision | QNN_CONTEXT_BINARY | float | Qualcomm® Dragonwing™ IQ-9075 | 989.927 ms | 11 - 34 MB | NPU |
| vision | QNN_CONTEXT_BINARY | float | Qualcomm® Dragonwing™ IQ-X7181 | 958.064 ms | 11 - 11 MB | NPU |
| vision | QNN_CONTEXT_BINARY | float | Qualcomm® Dragonwing™ Q-8750 | 527.153 ms | 11 - 22 MB | NPU |
| vision | QNN_CONTEXT_BINARY | float | Qualcomm® SA8295P | 1153.389 ms | 0 - 5 MB | NPU |
License
- The license for the original implementation of OWL-V2 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.