Deep Layer Aggregation
Paper • 1707.06484 • Published
DLA-102-X: Optimized for Qualcomm Devices
dla is a machine learning model that can classify images from the Imagenet dataset. It is a architecture which can be used as backbone.
This is based on the implementation of DLA-102-X 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 |
|---|---|---|---|---|
| ONNX | float | Universal | QAIRT 2.45, ONNX Runtime 1.25.0 | Download |
| ONNX | w8a8 | Universal | QAIRT 2.45, ONNX Runtime 1.25.0 | Download |
| QNN_DLC | float | Universal | QAIRT 2.45 | Download |
| QNN_DLC | w8a8 | Universal | QAIRT 2.45 | Download |
| TFLITE | float | Universal | QAIRT 2.45 | Download |
| TFLITE | w8a8 | Universal | QAIRT 2.45 | Download |
For more device-specific assets and performance metrics, visit DLA-102-X 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 DLA-102-X on GitHub for usage instructions.
Model Details
Model Type: Model_use_case.image_classification
Model Stats:
- Model checkpoint: Imagenet
- Input resolution: 224x224
- Number of parameters: 26.3M
- Model size (float): 100 MB
Performance Summary
| Model | Runtime | Precision | Chipset | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit |
|---|---|---|---|---|---|---|
| DLA-102-X | ONNX | float | Snapdragon® X2 Elite | 1.328 ms | 212 - 212 MB | NPU |
| DLA-102-X | ONNX | float | Snapdragon® X Elite | 2.698 ms | 148 - 148 MB | NPU |
| DLA-102-X | ONNX | float | Snapdragon® 8 Gen 3 Mobile | 1.868 ms | 1 - 144 MB | NPU |
| DLA-102-X | ONNX | float | Snapdragon® 8 Gen 1 Mobile | 3.925 ms | 1 - 119 MB | NPU |
| DLA-102-X | ONNX | float | Qualcomm® QCS8550 (Proxy) | 2.716 ms | 0 - 59 MB | NPU |
| DLA-102-X | ONNX | float | Qualcomm® QCS8450 | 3.925 ms | 1 - 119 MB | NPU |
| DLA-102-X | ONNX | float | Snapdragon® 8 Elite Mobile | 1.57 ms | 0 - 102 MB | NPU |
| DLA-102-X | ONNX | float | Snapdragon® 8 Elite Gen 5 Mobile | 1.219 ms | 1 - 99 MB | NPU |
| DLA-102-X | ONNX | float | Qualcomm® QCS9075 | 3.968 ms | 1 - 46 MB | NPU |
| DLA-102-X | ONNX | float | Qualcomm® QCS8750 | 1.57 ms | 0 - 102 MB | NPU |
| DLA-102-X | ONNX | float | Qualcomm® QCS7181 | 2.698 ms | 148 - 148 MB | NPU |
| DLA-102-X | ONNX | w8a8 | Snapdragon® X2 Elite | 0.719 ms | 213 - 213 MB | NPU |
| DLA-102-X | ONNX | w8a8 | Snapdragon® X Elite | 1.413 ms | 181 - 181 MB | NPU |
| DLA-102-X | ONNX | w8a8 | Snapdragon® 8 Gen 3 Mobile | 0.996 ms | 0 - 142 MB | NPU |
| DLA-102-X | ONNX | w8a8 | Snapdragon® 8 Gen 1 Mobile | 1.962 ms | 0 - 141 MB | NPU |
| DLA-102-X | ONNX | w8a8 | Qualcomm® QCS6490 | 4.013 ms | 0 - 45 MB | NPU |
| DLA-102-X | ONNX | w8a8 | Qualcomm® QCS8550 (Proxy) | 1.369 ms | 0 - 33 MB | NPU |
| DLA-102-X | ONNX | w8a8 | Qualcomm® QCS8450 | 1.962 ms | 0 - 141 MB | NPU |
| DLA-102-X | ONNX | w8a8 | Snapdragon® 8 Elite Gen 5 Mobile | 0.733 ms | 0 - 113 MB | NPU |
| DLA-102-X | ONNX | w8a8 | Snapdragon® 7 Gen 4 Mobile | 1.676 ms | 0 - 109 MB | NPU |
| DLA-102-X | ONNX | w8a8 | Qualcomm® QCM6690 | 10.917 ms | 0 - 222 MB | NPU |
| DLA-102-X | ONNX | w8a8 | Qualcomm® QCS9075 | 1.634 ms | 0 - 45 MB | NPU |
| DLA-102-X | ONNX | w8a8 | Snapdragon® 8 Elite Mobile | 0.876 ms | 0 - 115 MB | NPU |
| DLA-102-X | ONNX | w8a8 | Qualcomm® QCS7790 | 1.676 ms | 0 - 109 MB | NPU |
| DLA-102-X | ONNX | w8a8 | Qualcomm® QCS8750 | 0.876 ms | 0 - 115 MB | NPU |
| DLA-102-X | ONNX | w8a8 | Qualcomm® QCS7181 | 1.413 ms | 181 - 181 MB | NPU |
| DLA-102-X | QNN_DLC | float | Snapdragon® X2 Elite | 1.628 ms | 1 - 1 MB | NPU |
| DLA-102-X | QNN_DLC | float | Snapdragon® X Elite | 3.133 ms | 1 - 1 MB | NPU |
| DLA-102-X | QNN_DLC | float | Snapdragon® 8 Gen 3 Mobile | 2.005 ms | 0 - 146 MB | NPU |
| DLA-102-X | QNN_DLC | float | Snapdragon® 8 Gen 1 Mobile | 4.551 ms | 0 - 122 MB | NPU |
| DLA-102-X | QNN_DLC | float | Qualcomm® QCS8275 | 14.436 ms | 1 - 99 MB | NPU |
| DLA-102-X | QNN_DLC | float | Qualcomm® QCS8550 (Proxy) | 2.868 ms | 1 - 3 MB | NPU |
| DLA-102-X | QNN_DLC | float | Qualcomm® QCS8450 | 4.551 ms | 0 - 122 MB | NPU |
| DLA-102-X | QNN_DLC | float | Snapdragon® 8 Elite Mobile | 1.603 ms | 1 - 100 MB | NPU |
| DLA-102-X | QNN_DLC | float | Qualcomm® SA8295P | 4.641 ms | 1 - 81 MB | NPU |
| DLA-102-X | QNN_DLC | float | Snapdragon® 8 Elite Gen 5 Mobile | 1.233 ms | 0 - 103 MB | NPU |
| DLA-102-X | QNN_DLC | float | Qualcomm® SA7255P | 14.436 ms | 1 - 99 MB | NPU |
| DLA-102-X | QNN_DLC | float | Qualcomm® QCS9075 | 4.213 ms | 3 - 5 MB | NPU |
| DLA-102-X | QNN_DLC | float | Qualcomm® QCS8750 | 1.603 ms | 1 - 100 MB | NPU |
| DLA-102-X | QNN_DLC | float | Qualcomm® QCS7181 | 3.133 ms | 1 - 1 MB | NPU |
| DLA-102-X | QNN_DLC | w8a8 | Snapdragon® X2 Elite | 0.871 ms | 0 - 0 MB | NPU |
| DLA-102-X | QNN_DLC | w8a8 | Snapdragon® X Elite | 1.583 ms | 1 - 1 MB | NPU |
| DLA-102-X | QNN_DLC | w8a8 | Snapdragon® 8 Gen 3 Mobile | 0.962 ms | 0 - 132 MB | NPU |
| DLA-102-X | QNN_DLC | w8a8 | Snapdragon® 8 Gen 1 Mobile | 1.933 ms | 0 - 135 MB | NPU |
| DLA-102-X | QNN_DLC | w8a8 | Qualcomm® QCS6490 | 3.942 ms | 2 - 4 MB | NPU |
| DLA-102-X | QNN_DLC | w8a8 | Qualcomm® QCS8275 | 3.235 ms | 0 - 105 MB | NPU |
| DLA-102-X | QNN_DLC | w8a8 | Qualcomm® QCS8550 (Proxy) | 1.407 ms | 0 - 2 MB | NPU |
| DLA-102-X | QNN_DLC | w8a8 | Qualcomm® QCS8450 | 1.933 ms | 0 - 135 MB | NPU |
| DLA-102-X | QNN_DLC | w8a8 | Snapdragon® 8 Elite Gen 5 Mobile | 0.631 ms | 0 - 109 MB | NPU |
| DLA-102-X | QNN_DLC | w8a8 | Snapdragon® 7 Gen 4 Mobile | 1.696 ms | 0 - 103 MB | NPU |
| DLA-102-X | QNN_DLC | w8a8 | Qualcomm® QCM6690 | 10.647 ms | 0 - 236 MB | NPU |
| DLA-102-X | QNN_DLC | w8a8 | Qualcomm® QCS9075 | 1.581 ms | 0 - 2 MB | NPU |
| DLA-102-X | QNN_DLC | w8a8 | Qualcomm® SA7255P | 3.235 ms | 0 - 105 MB | NPU |
| DLA-102-X | QNN_DLC | w8a8 | Snapdragon® 8 Elite Mobile | 0.791 ms | 0 - 108 MB | NPU |
| DLA-102-X | QNN_DLC | w8a8 | Qualcomm® SA8295P | 2.156 ms | 0 - 108 MB | NPU |
| DLA-102-X | QNN_DLC | w8a8 | Qualcomm® QCS7790 | 1.696 ms | 0 - 103 MB | NPU |
| DLA-102-X | QNN_DLC | w8a8 | Qualcomm® QCS8750 | 0.791 ms | 0 - 108 MB | NPU |
| DLA-102-X | QNN_DLC | w8a8 | Qualcomm® QCS7181 | 1.583 ms | 1 - 1 MB | NPU |
| DLA-102-X | TFLITE | float | Snapdragon® 8 Gen 3 Mobile | 1.979 ms | 0 - 195 MB | NPU |
| DLA-102-X | TFLITE | float | Snapdragon® 8 Gen 1 Mobile | 4.521 ms | 0 - 166 MB | NPU |
| DLA-102-X | TFLITE | float | Qualcomm® QCS8275 | 14.415 ms | 0 - 145 MB | NPU |
| DLA-102-X | TFLITE | float | Qualcomm® QCS8550 (Proxy) | 2.737 ms | 0 - 2 MB | NPU |
| DLA-102-X | TFLITE | float | Qualcomm® SA8775P | 29.37 ms | 0 - 23 MB | GPU |
| DLA-102-X | TFLITE | float | Qualcomm® SA8650P | 29.37 ms | 0 - 23 MB | GPU |
| DLA-102-X | TFLITE | float | Qualcomm® SA8255P | 29.37 ms | 0 - 23 MB | GPU |
| DLA-102-X | TFLITE | float | Qualcomm® QCS8450 | 4.521 ms | 0 - 166 MB | NPU |
| DLA-102-X | TFLITE | float | Snapdragon® 8 Elite Mobile | 1.581 ms | 0 - 146 MB | NPU |
| DLA-102-X | TFLITE | float | Qualcomm® SA8295P | 4.663 ms | 0 - 125 MB | NPU |
| DLA-102-X | TFLITE | float | Snapdragon® 8 Elite Gen 5 Mobile | 1.227 ms | 0 - 152 MB | NPU |
| DLA-102-X | TFLITE | float | Qualcomm® SA7255P | 14.415 ms | 0 - 145 MB | NPU |
| DLA-102-X | TFLITE | float | Qualcomm® QCS9075 | 4.244 ms | 0 - 57 MB | NPU |
| DLA-102-X | TFLITE | float | Qualcomm® QCS8750 | 1.581 ms | 0 - 146 MB | NPU |
| DLA-102-X | TFLITE | w8a8 | Snapdragon® 8 Gen 3 Mobile | 0.817 ms | 0 - 135 MB | NPU |
| DLA-102-X | TFLITE | w8a8 | Snapdragon® 8 Gen 1 Mobile | 1.74 ms | 0 - 135 MB | NPU |
| DLA-102-X | TFLITE | w8a8 | Qualcomm® QCS6490 | 3.548 ms | 0 - 29 MB | NPU |
| DLA-102-X | TFLITE | w8a8 | Qualcomm® QCS8275 | 2.972 ms | 0 - 99 MB | NPU |
| DLA-102-X | TFLITE | w8a8 | Qualcomm® QCS8550 (Proxy) | 1.187 ms | 0 - 2 MB | NPU |
| DLA-102-X | TFLITE | w8a8 | Qualcomm® SA8775P | 30.549 ms | 1 - 26 MB | GPU |
| DLA-102-X | TFLITE | w8a8 | Qualcomm® SA8650P | 30.549 ms | 1 - 26 MB | GPU |
| DLA-102-X | TFLITE | w8a8 | Qualcomm® SA8255P | 30.549 ms | 1 - 26 MB | GPU |
| DLA-102-X | TFLITE | w8a8 | Qualcomm® QCS8450 | 1.74 ms | 0 - 135 MB | NPU |
| DLA-102-X | TFLITE | w8a8 | Snapdragon® 8 Elite Gen 5 Mobile | 0.568 ms | 0 - 95 MB | NPU |
| DLA-102-X | TFLITE | w8a8 | Snapdragon® 7 Gen 4 Mobile | 1.474 ms | 0 - 92 MB | NPU |
| DLA-102-X | TFLITE | w8a8 | Qualcomm® QCM6690 | 10.242 ms | 0 - 208 MB | NPU |
| DLA-102-X | TFLITE | w8a8 | Qualcomm® QCS9075 | 1.361 ms | 0 - 29 MB | NPU |
| DLA-102-X | TFLITE | w8a8 | Qualcomm® SA7255P | 2.972 ms | 0 - 99 MB | NPU |
| DLA-102-X | TFLITE | w8a8 | Snapdragon® 8 Elite Mobile | 0.712 ms | 0 - 105 MB | NPU |
| DLA-102-X | TFLITE | w8a8 | Qualcomm® SA8295P | 1.946 ms | 0 - 103 MB | NPU |
| DLA-102-X | TFLITE | w8a8 | Qualcomm® QCS7790 | 1.474 ms | 0 - 92 MB | NPU |
| DLA-102-X | TFLITE | w8a8 | Qualcomm® QCS8750 | 0.712 ms | 0 - 105 MB | NPU |
License
- The license for the original implementation of DLA-102-X 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.