metadata
library_name: pytorch
license: other
tags:
- android
pipeline_tag: image-to-image
Real-ESRGAN-x4plus: Optimized for Qualcomm Devices
Real-ESRGAN is a machine learning model that upscales an image with minimal loss in quality. The implementation is a derivative of the Real-ESRGAN-x4plus architecture, a larger and more powerful version compared to the Real-ESRGAN-general-x4v3 architecture.
This is based on the implementation of Real-ESRGAN-x4plus 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 2.42, ONNX Runtime 1.24.3 | Download |
| ONNX | w8a8 | Universal | QAIRT 2.42, ONNX Runtime 1.24.3 | 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 Real-ESRGAN-x4plus 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 Real-ESRGAN-x4plus on GitHub for usage instructions.
Model Details
Model Type: Model_use_case.super_resolution
Model Stats:
- Model checkpoint: RealESRGAN_x4plus
- Input resolution: 128x128
- Number of parameters: 16.7M
- Model size (float): 63.9 MB
- Model size (w8a8): 16.7 MB
Performance Summary
| Model | Runtime | Precision | Chipset | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit |
|---|---|---|---|---|---|---|
| Real-ESRGAN-x4plus | ONNX | float | Snapdragon® 8 Elite Gen 5 Mobile | 26.646 ms | 7 - 353 MB | NPU |
| Real-ESRGAN-x4plus | ONNX | float | Snapdragon® 8 Elite Mobile | 38.148 ms | 7 - 351 MB | NPU |
| Real-ESRGAN-x4plus | ONNX | float | Snapdragon® X2 Elite | 34.545 ms | 37 - 37 MB | NPU |
| Real-ESRGAN-x4plus | ONNX | float | Snapdragon® X Elite | 65.579 ms | 37 - 37 MB | NPU |
| Real-ESRGAN-x4plus | ONNX | float | Snapdragon® X Elite | 65.579 ms | 37 - 37 MB | NPU |
| Real-ESRGAN-x4plus | ONNX | float | Snapdragon® 8 Gen 3 Mobile | 51.64 ms | 9 - 801 MB | NPU |
| Real-ESRGAN-x4plus | ONNX | float | Qualcomm® QCS8550 (Proxy) | 69.166 ms | 0 - 44 MB | NPU |
| Real-ESRGAN-x4plus | ONNX | float | Snapdragon® 8 Elite For Galaxy Mobile | 38.148 ms | 7 - 351 MB | NPU |
| Real-ESRGAN-x4plus | ONNX | float | Qualcomm® QCS9075 | 107.407 ms | 6 - 9 MB | NPU |
| Real-ESRGAN-x4plus | ONNX | w8a8 | Snapdragon® 8 Elite Gen 5 Mobile | 9.122 ms | 2 - 498 MB | NPU |
| Real-ESRGAN-x4plus | ONNX | w8a8 | Snapdragon® 8 Elite Mobile | 12.63 ms | 2 - 487 MB | NPU |
| Real-ESRGAN-x4plus | ONNX | w8a8 | Snapdragon® X2 Elite | 10.339 ms | 23 - 23 MB | NPU |
| Real-ESRGAN-x4plus | ONNX | w8a8 | Snapdragon® X Elite | 26.232 ms | 21 - 21 MB | NPU |
| Real-ESRGAN-x4plus | ONNX | w8a8 | Snapdragon® X Elite | 26.232 ms | 21 - 21 MB | NPU |
| Real-ESRGAN-x4plus | ONNX | w8a8 | Snapdragon® 8 Gen 3 Mobile | 17.001 ms | 2 - 786 MB | NPU |
| Real-ESRGAN-x4plus | ONNX | w8a8 | Qualcomm® QCS8550 (Proxy) | 24.935 ms | 0 - 30 MB | NPU |
| Real-ESRGAN-x4plus | ONNX | w8a8 | Qualcomm® QCS9075 | 30.533 ms | 2 - 4 MB | NPU |
| Real-ESRGAN-x4plus | ONNX | w8a8 | Snapdragon® 8 Elite For Galaxy Mobile | 12.63 ms | 2 - 487 MB | NPU |
| Real-ESRGAN-x4plus | QNN_DLC | float | Snapdragon® 8 Elite Gen 5 Mobile | 25.145 ms | 0 - 331 MB | NPU |
| Real-ESRGAN-x4plus | QNN_DLC | float | Snapdragon® 8 Elite Mobile | 37.735 ms | 0 - 334 MB | NPU |
| Real-ESRGAN-x4plus | QNN_DLC | float | Snapdragon® X2 Elite | 34.501 ms | 0 - 0 MB | NPU |
| Real-ESRGAN-x4plus | QNN_DLC | float | Snapdragon® X Elite | 64.972 ms | 0 - 0 MB | NPU |
| Real-ESRGAN-x4plus | QNN_DLC | float | Snapdragon® X Elite | 64.972 ms | 0 - 0 MB | NPU |
| Real-ESRGAN-x4plus | QNN_DLC | float | Snapdragon® 8 Gen 3 Mobile | 48.622 ms | 0 - 707 MB | NPU |
| Real-ESRGAN-x4plus | QNN_DLC | float | Qualcomm® QCS8550 (Proxy) | 64.803 ms | 0 - 4 MB | NPU |
| Real-ESRGAN-x4plus | QNN_DLC | float | Qualcomm® SA8775P | 105.552 ms | 0 - 353 MB | NPU |
| Real-ESRGAN-x4plus | QNN_DLC | float | Qualcomm® SA8775P | 105.552 ms | 0 - 353 MB | NPU |
| Real-ESRGAN-x4plus | QNN_DLC | float | Qualcomm® SA8775P | 105.552 ms | 0 - 353 MB | NPU |
| Real-ESRGAN-x4plus | QNN_DLC | float | Qualcomm® SA7255P | 452.107 ms | 0 - 352 MB | NPU |
| Real-ESRGAN-x4plus | QNN_DLC | float | Qualcomm® QCS8450 (Proxy) | 112.4 ms | 1 - 710 MB | NPU |
| Real-ESRGAN-x4plus | QNN_DLC | float | Qualcomm® SA8295P | 111.272 ms | 0 - 359 MB | NPU |
| Real-ESRGAN-x4plus | QNN_DLC | float | Snapdragon® 8 Elite For Galaxy Mobile | 37.735 ms | 0 - 334 MB | NPU |
| Real-ESRGAN-x4plus | QNN_DLC | float | Qualcomm® QCS9075 | 108.883 ms | 0 - 5 MB | NPU |
| Real-ESRGAN-x4plus | QNN_DLC | w8a8 | Snapdragon® 8 Elite Gen 5 Mobile | 9.079 ms | 0 - 436 MB | NPU |
| Real-ESRGAN-x4plus | QNN_DLC | w8a8 | Snapdragon® 8 Elite Mobile | 12.298 ms | 0 - 424 MB | NPU |
| Real-ESRGAN-x4plus | QNN_DLC | w8a8 | Snapdragon® X2 Elite | 10.385 ms | 0 - 0 MB | NPU |
| Real-ESRGAN-x4plus | QNN_DLC | w8a8 | Snapdragon® X Elite | 25.877 ms | 0 - 0 MB | NPU |
| Real-ESRGAN-x4plus | QNN_DLC | w8a8 | Snapdragon® X Elite | 25.877 ms | 0 - 0 MB | NPU |
| Real-ESRGAN-x4plus | QNN_DLC | w8a8 | Snapdragon® 8 Gen 3 Mobile | 16.774 ms | 0 - 687 MB | NPU |
| Real-ESRGAN-x4plus | QNN_DLC | w8a8 | Qualcomm® QCS6490 | 110.031 ms | 0 - 2 MB | NPU |
| Real-ESRGAN-x4plus | QNN_DLC | w8a8 | Qualcomm® QCS8550 (Proxy) | 24.502 ms | 0 - 4 MB | NPU |
| Real-ESRGAN-x4plus | QNN_DLC | w8a8 | Qualcomm® SA8775P | 22.828 ms | 0 - 486 MB | NPU |
| Real-ESRGAN-x4plus | QNN_DLC | w8a8 | Qualcomm® SA8775P | 22.828 ms | 0 - 486 MB | NPU |
| Real-ESRGAN-x4plus | QNN_DLC | w8a8 | Qualcomm® SA8775P | 22.828 ms | 0 - 486 MB | NPU |
| Real-ESRGAN-x4plus | QNN_DLC | w8a8 | Qualcomm® QCS9075 | 31.104 ms | 0 - 2 MB | NPU |
| Real-ESRGAN-x4plus | QNN_DLC | w8a8 | Snapdragon® 8 Elite For Galaxy Mobile | 12.298 ms | 0 - 424 MB | NPU |
| Real-ESRGAN-x4plus | QNN_DLC | w8a8 | Qualcomm® QCS8450 (Proxy) | 39.499 ms | 0 - 719 MB | NPU |
| Real-ESRGAN-x4plus | QNN_DLC | w8a8 | Qualcomm® SA7255P | 68.994 ms | 0 - 486 MB | NPU |
| Real-ESRGAN-x4plus | QNN_DLC | w8a8 | Snapdragon® 7 Gen 4 Mobile | 38.61 ms | 0 - 487 MB | NPU |
| Real-ESRGAN-x4plus | QNN_DLC | w8a8 | Qualcomm® QCM6690 | 481.354 ms | 0 - 462 MB | NPU |
| Real-ESRGAN-x4plus | QNN_DLC | w8a8 | Qualcomm® SA8295P | 36.425 ms | 0 - 496 MB | NPU |
| Real-ESRGAN-x4plus | QNN_DLC | w8a8 | Snapdragon® 7 Gen 4 Mobile | 38.61 ms | 0 - 487 MB | NPU |
| Real-ESRGAN-x4plus | TFLITE | float | Snapdragon® 8 Elite Gen 5 Mobile | 26.12 ms | 3 - 370 MB | NPU |
| Real-ESRGAN-x4plus | TFLITE | float | Snapdragon® 8 Elite Mobile | 37.651 ms | 3 - 366 MB | NPU |
| Real-ESRGAN-x4plus | TFLITE | float | Snapdragon® 8 Gen 3 Mobile | 49.516 ms | 3 - 749 MB | NPU |
| Real-ESRGAN-x4plus | TFLITE | float | Qualcomm® QCS8550 (Proxy) | 63.403 ms | 3 - 6 MB | NPU |
| Real-ESRGAN-x4plus | TFLITE | float | Qualcomm® SA8775P | 105.883 ms | 3 - 394 MB | NPU |
| Real-ESRGAN-x4plus | TFLITE | float | Qualcomm® SA8775P | 105.883 ms | 3 - 394 MB | NPU |
| Real-ESRGAN-x4plus | TFLITE | float | Qualcomm® SA8775P | 105.883 ms | 3 - 394 MB | NPU |
| Real-ESRGAN-x4plus | TFLITE | float | Qualcomm® SA7255P | 452.665 ms | 3 - 394 MB | NPU |
| Real-ESRGAN-x4plus | TFLITE | float | Qualcomm® QCS8450 (Proxy) | 115.589 ms | 3 - 731 MB | NPU |
| Real-ESRGAN-x4plus | TFLITE | float | Qualcomm® SA8295P | 110.485 ms | 3 - 391 MB | NPU |
| Real-ESRGAN-x4plus | TFLITE | float | Snapdragon® 8 Elite For Galaxy Mobile | 37.651 ms | 3 - 366 MB | NPU |
| Real-ESRGAN-x4plus | TFLITE | float | Qualcomm® QCS9075 | 109.563 ms | 1 - 45 MB | NPU |
| Real-ESRGAN-x4plus | TFLITE | w8a8 | Snapdragon® 8 Elite Gen 5 Mobile | 7.607 ms | 1 - 506 MB | NPU |
| Real-ESRGAN-x4plus | TFLITE | w8a8 | Snapdragon® 8 Elite Mobile | 11.705 ms | 1 - 480 MB | NPU |
| Real-ESRGAN-x4plus | TFLITE | w8a8 | Snapdragon® 8 Gen 3 Mobile | 13.888 ms | 1 - 654 MB | NPU |
| Real-ESRGAN-x4plus | TFLITE | w8a8 | Qualcomm® QCS6490 | 93.625 ms | 1 - 26 MB | NPU |
| Real-ESRGAN-x4plus | TFLITE | w8a8 | Qualcomm® QCS8550 (Proxy) | 19.885 ms | 1 - 4 MB | NPU |
| Real-ESRGAN-x4plus | TFLITE | w8a8 | Qualcomm® SA8775P | 18.266 ms | 1 - 451 MB | NPU |
| Real-ESRGAN-x4plus | TFLITE | w8a8 | Qualcomm® SA8775P | 18.266 ms | 1 - 451 MB | NPU |
| Real-ESRGAN-x4plus | TFLITE | w8a8 | Qualcomm® SA8775P | 18.266 ms | 1 - 451 MB | NPU |
| Real-ESRGAN-x4plus | TFLITE | w8a8 | Qualcomm® QCS9075 | 25.437 ms | 0 - 25 MB | NPU |
| Real-ESRGAN-x4plus | TFLITE | w8a8 | Snapdragon® 8 Elite For Galaxy Mobile | 11.705 ms | 1 - 480 MB | NPU |
| Real-ESRGAN-x4plus | TFLITE | w8a8 | Qualcomm® QCS8450 (Proxy) | 35.668 ms | 1 - 664 MB | NPU |
| Real-ESRGAN-x4plus | TFLITE | w8a8 | Qualcomm® SA7255P | 64.858 ms | 1 - 448 MB | NPU |
| Real-ESRGAN-x4plus | TFLITE | w8a8 | Snapdragon® 7 Gen 4 Mobile | 33.046 ms | 1 - 449 MB | NPU |
| Real-ESRGAN-x4plus | TFLITE | w8a8 | Qualcomm® QCM6690 | 439.371 ms | 1 - 437 MB | NPU |
| Real-ESRGAN-x4plus | TFLITE | w8a8 | Qualcomm® SA8295P | 34.543 ms | 1 - 447 MB | NPU |
| Real-ESRGAN-x4plus | TFLITE | w8a8 | Snapdragon® 7 Gen 4 Mobile | 33.046 ms | 1 - 449 MB | NPU |
License
- The license for the original implementation of Real-ESRGAN-x4plus can be found here.
References
- Real-ESRGAN: Training Real-World Blind Super-Resolution with Pure Synthetic Data
- 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.