ESRGAN: Optimized for Qualcomm Devices

ESRGAN is a machine learning model that upscales an image with minimal loss in quality.

This is based on the implementation of ESRGAN 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 2.37, ONNX Runtime 1.23.0	Download
QNN_DLC	float	Universal	QAIRT 2.42	Download
TFLITE	float	Universal	QAIRT 2.42, TFLite 2.17.0	Download

For more device-specific assets and performance metrics, visit ESRGAN 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 ESRGAN on GitHub for usage instructions.

Model Details

Model Type: Model_use_case.super_resolution

Model Stats:

• Model checkpoint: ESRGAN_x4
• Input resolution: 128x128
• Number of parameters: 16.7M
• Model size (float): 63.9 MB

Performance Summary

Model	Runtime	Precision	Chipset	Inference Time (ms)	Peak Memory Range (MB)	Primary Compute Unit
ESRGAN	ONNX	float	Snapdragon® X Elite	65.559 ms	38 - 38 MB	NPU
ESRGAN	ONNX	float	Snapdragon® 8 Gen 3 Mobile	51.0 ms	7 - 697 MB	NPU
ESRGAN	ONNX	float	Qualcomm® QCS8550 (Proxy)	65.676 ms	0 - 45 MB	NPU
ESRGAN	ONNX	float	Qualcomm® QCS9075	106.901 ms	6 - 9 MB	NPU
ESRGAN	ONNX	float	Snapdragon® 8 Elite For Galaxy Mobile	38.988 ms	7 - 257 MB	NPU
ESRGAN	ONNX	float	Snapdragon® 8 Elite Gen 5 Mobile	28.071 ms	0 - 261 MB	NPU
ESRGAN	QNN_DLC	float	Snapdragon® X Elite	64.947 ms	0 - 0 MB	NPU
ESRGAN	QNN_DLC	float	Snapdragon® 8 Gen 3 Mobile	49.399 ms	0 - 748 MB	NPU
ESRGAN	QNN_DLC	float	Qualcomm® QCS8275 (Proxy)	451.839 ms	0 - 347 MB	NPU
ESRGAN	QNN_DLC	float	Qualcomm® QCS8550 (Proxy)	62.453 ms	0 - 3 MB	NPU
ESRGAN	QNN_DLC	float	Qualcomm® SA8775P	105.38 ms	0 - 347 MB	NPU
ESRGAN	QNN_DLC	float	Qualcomm® QCS9075	106.848 ms	0 - 5 MB	NPU
ESRGAN	QNN_DLC	float	Qualcomm® QCS8450 (Proxy)	112.556 ms	0 - 753 MB	NPU
ESRGAN	QNN_DLC	float	Qualcomm® SA7255P	451.839 ms	0 - 347 MB	NPU
ESRGAN	QNN_DLC	float	Qualcomm® SA8295P	111.274 ms	0 - 357 MB	NPU
ESRGAN	QNN_DLC	float	Snapdragon® 8 Elite For Galaxy Mobile	38.15 ms	0 - 332 MB	NPU
ESRGAN	QNN_DLC	float	Snapdragon® 8 Elite Gen 5 Mobile	25.883 ms	0 - 331 MB	NPU
ESRGAN	TFLITE	float	Snapdragon® 8 Gen 3 Mobile	49.641 ms	3 - 798 MB	NPU
ESRGAN	TFLITE	float	Qualcomm® QCS8275 (Proxy)	451.971 ms	7 - 398 MB	NPU
ESRGAN	TFLITE	float	Qualcomm® QCS8550 (Proxy)	59.589 ms	3 - 6 MB	NPU
ESRGAN	TFLITE	float	Qualcomm® SA8775P	105.469 ms	3 - 394 MB	NPU
ESRGAN	TFLITE	float	Qualcomm® QCS9075	107.96 ms	3 - 47 MB	NPU
ESRGAN	TFLITE	float	Qualcomm® QCS8450 (Proxy)	122.167 ms	3 - 797 MB	NPU
ESRGAN	TFLITE	float	Qualcomm® SA7255P	451.971 ms	7 - 398 MB	NPU
ESRGAN	TFLITE	float	Qualcomm® SA8295P	111.299 ms	3 - 398 MB	NPU
ESRGAN	TFLITE	float	Snapdragon® 8 Elite For Galaxy Mobile	38.09 ms	3 - 376 MB	NPU
ESRGAN	TFLITE	float	Snapdragon® 8 Elite Gen 5 Mobile	27.14 ms	1 - 374 MB	NPU

License

• The license for the original implementation of ESRGAN can be found here.

References

• ESRGAN: Enhanced Super-Resolution Generative Adversarial Networks
• 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.