NAFNet-DeNoise: Optimized for Qualcomm Devices

NAFNET is designed for lightweight real-time denoising of images.

This is based on the implementation of NAFNet-DeNoise 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 2.50, ONNX Runtime 1.27.1	Download
ONNX	w8a16	Universal	QAIRT 2.50, ONNX Runtime 1.27.1	Download
QNN_DLC	w8a16	Universal	QAIRT 2.50	Download

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

Model Details

Model Type: Model_use_case.image_editing

Model Stats:

• Input resolution: 256x256
• Model checkpoint: NAFNet-SIDD-width64
• Model size (float): 463.93 MB
• Number of parameters: 115.98M

Performance Summary

Model	Runtime	Precision	Chipset	Inference Time (ms)	Peak Memory Range (MB)	Primary Compute Unit
NAFNet-DeNoise	ONNX	float	Snapdragon® 8 Elite Gen 5 For Galaxy Mobile	14.789 ms	2 - 494 MB	NPU
NAFNet-DeNoise	ONNX	float	Snapdragon® 8 Elite For Galaxy Mobile	18.051 ms	2 - 455 MB	NPU
NAFNet-DeNoise	ONNX	float	Snapdragon® X2 Elite	14.901 ms	4 - 4 MB	NPU
NAFNet-DeNoise	ONNX	float	Snapdragon® X Elite	31.339 ms	227 - 227 MB	NPU
NAFNet-DeNoise	ONNX	float	Snapdragon® 8 Gen 3 Mobile	22.352 ms	2 - 797 MB	NPU
NAFNet-DeNoise	ONNX	float	Snapdragon® 8 Gen 1 Mobile	48.902 ms	3 - 707 MB	NPU
NAFNet-DeNoise	ONNX	float	Qualcomm® Dragonwing™ IQ-8275	40.924 ms	2 - 6 MB	NPU
NAFNet-DeNoise	ONNX	float	Qualcomm® Dragonwing™ QCS8550 (Proxy)	31.024 ms	0 - 258 MB	NPU
NAFNet-DeNoise	ONNX	float	Qualcomm® QCS8450	48.902 ms	3 - 707 MB	NPU
NAFNet-DeNoise	ONNX	float	Qualcomm® Dragonwing™ IQ-9075	39.289 ms	2 - 6 MB	NPU
NAFNet-DeNoise	ONNX	float	Qualcomm® Dragonwing™ IQ-X7181	31.339 ms	227 - 227 MB	NPU
NAFNet-DeNoise	ONNX	float	Qualcomm® Dragonwing™ Q-8750	18.051 ms	2 - 455 MB	NPU
NAFNet-DeNoise	ONNX	w8a16	Snapdragon® 8 Elite Gen 5 For Galaxy Mobile	10.562 ms	0 - 508 MB	NPU
NAFNet-DeNoise	ONNX	w8a16	Snapdragon® 8 Elite For Galaxy Mobile	16.309 ms	1 - 452 MB	NPU
NAFNet-DeNoise	ONNX	w8a16	Snapdragon® X2 Elite	11.925 ms	1 - 1 MB	NPU
NAFNet-DeNoise	ONNX	w8a16	Snapdragon® X Elite	28.346 ms	118 - 118 MB	NPU
NAFNet-DeNoise	ONNX	w8a16	Snapdragon® 8 Gen 3 Mobile	20.865 ms	0 - 574 MB	NPU
NAFNet-DeNoise	ONNX	w8a16	Snapdragon® 8 Gen 1 Mobile	37.324 ms	0 - 589 MB	NPU
NAFNet-DeNoise	ONNX	w8a16	Qualcomm® Dragonwing™ QCS6490	86.855 ms	1 - 4 MB	NPU
NAFNet-DeNoise	ONNX	w8a16	Qualcomm® Dragonwing™ IQ-8275	26.686 ms	1 - 5 MB	NPU
NAFNet-DeNoise	ONNX	w8a16	Qualcomm® Dragonwing™ QCS8550 (Proxy)	28.476 ms	0 - 131 MB	NPU
NAFNet-DeNoise	ONNX	w8a16	Qualcomm® QCS8450	37.324 ms	0 - 589 MB	NPU
NAFNet-DeNoise	ONNX	w8a16	Qualcomm® Dragonwing™ IQ-9075	27.425 ms	1 - 4 MB	NPU
NAFNet-DeNoise	ONNX	w8a16	Qualcomm® Dragonwing™ IQ-X7181	28.346 ms	118 - 118 MB	NPU
NAFNet-DeNoise	ONNX	w8a16	Qualcomm® Dragonwing™ Q-6690	298.804 ms	1 - 675 MB	NPU
NAFNet-DeNoise	ONNX	w8a16	Qualcomm® Dragonwing™ Q-7790	42.925 ms	1 - 524 MB	NPU
NAFNet-DeNoise	ONNX	w8a16	Qualcomm® Dragonwing™ Q-8750	16.309 ms	1 - 452 MB	NPU
NAFNet-DeNoise	ONNX	w8a16	Snapdragon® 7 Gen 4 Mobile	42.925 ms	1 - 524 MB	NPU
NAFNet-DeNoise	QNN_DLC	w8a16	Snapdragon® 8 Elite Gen 5 For Galaxy Mobile	11.211 ms	0 - 494 MB	NPU
NAFNet-DeNoise	QNN_DLC	w8a16	Snapdragon® 8 Elite For Galaxy Mobile	15.701 ms	0 - 461 MB	NPU
NAFNet-DeNoise	QNN_DLC	w8a16	Snapdragon® X2 Elite	12.879 ms	0 - 0 MB	NPU
NAFNet-DeNoise	QNN_DLC	w8a16	Snapdragon® X Elite	28.657 ms	0 - 0 MB	NPU
NAFNet-DeNoise	QNN_DLC	w8a16	Snapdragon® 8 Gen 3 Mobile	20.997 ms	0 - 565 MB	NPU
NAFNet-DeNoise	QNN_DLC	w8a16	Snapdragon® 8 Gen 1 Mobile	38.079 ms	0 - 576 MB	NPU
NAFNet-DeNoise	QNN_DLC	w8a16	Qualcomm® Dragonwing™ QCS6490	86.656 ms	0 - 3 MB	NPU
NAFNet-DeNoise	QNN_DLC	w8a16	Qualcomm® Dragonwing™ IQ-8275	26.695 ms	0 - 4 MB	NPU
NAFNet-DeNoise	QNN_DLC	w8a16	Qualcomm® Dragonwing™ QCS8550 (Proxy)	28.474 ms	1 - 431 MB	NPU
NAFNet-DeNoise	QNN_DLC	w8a16	Qualcomm® SA8775P	26.743 ms	1 - 460 MB	NPU
NAFNet-DeNoise	QNN_DLC	w8a16	Qualcomm® SA8650P	26.743 ms	1 - 460 MB	NPU
NAFNet-DeNoise	QNN_DLC	w8a16	Qualcomm® SA8255P	26.743 ms	1 - 460 MB	NPU
NAFNet-DeNoise	QNN_DLC	w8a16	Qualcomm® QCS8450	38.079 ms	0 - 576 MB	NPU
NAFNet-DeNoise	QNN_DLC	w8a16	Qualcomm® Dragonwing™ IQ-9075	26.856 ms	0 - 3 MB	NPU
NAFNet-DeNoise	QNN_DLC	w8a16	Qualcomm® Dragonwing™ IQ-X7181	28.657 ms	0 - 0 MB	NPU
NAFNet-DeNoise	QNN_DLC	w8a16	Qualcomm® Dragonwing™ Q-6690	271.446 ms	1 - 632 MB	NPU
NAFNet-DeNoise	QNN_DLC	w8a16	Qualcomm® Dragonwing™ Q-7790	47.371 ms	0 - 549 MB	NPU
NAFNet-DeNoise	QNN_DLC	w8a16	Qualcomm® Dragonwing™ Q-8750	15.701 ms	0 - 461 MB	NPU
NAFNet-DeNoise	QNN_DLC	w8a16	Qualcomm® SA8295P	36.422 ms	1 - 459 MB	NPU
NAFNet-DeNoise	QNN_DLC	w8a16	Snapdragon® 7 Gen 4 Mobile	47.371 ms	0 - 549 MB	NPU

License

• The license for the original implementation of NAFNet-DeNoise can be found here.

References

• Simple Baselines for Image Restoration
• 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.