DnCNN: Optimized for Qualcomm Devices

DnCNN is a 17-layer denoising convolutional neural network that uses residual learning to remove Gaussian noise (sigma=25) from grayscale images. The network predicts the noise residual and subtracts it from the input to produce a clean image.

This is based on the implementation of DnCNN 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 DnCNN 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 DnCNN on GitHub for usage instructions.

Model Details

Model Type: Model_use_case.image_editing

Model Stats:

Model checkpoint: dncnn_25
Input resolution: 256x256
Number of parameters: 555K
Model size (float): 2.12 MB
Model size (w8a8): 581 KB

Performance Summary

Model	Runtime	Precision	Chipset	Inference Time (ms)	Peak Memory Range (MB)	Primary Compute Unit
DnCNN	ONNX	float	Snapdragon® X2 Elite	4.042 ms	213 - 213 MB	NPU
DnCNN	ONNX	float	Snapdragon® X Elite	7.137 ms	180 - 180 MB	NPU
DnCNN	ONNX	float	Snapdragon® 8 Gen 3 Mobile	5.17 ms	2 - 171 MB	NPU
DnCNN	ONNX	float	Snapdragon® 8 Gen 1 Mobile	13.831 ms	1 - 170 MB	NPU
DnCNN	ONNX	float	Qualcomm® QCS8550 (Proxy)	6.94 ms	1 - 4 MB	NPU
DnCNN	ONNX	float	Qualcomm® QCS8450	13.831 ms	1 - 170 MB	NPU
DnCNN	ONNX	float	Snapdragon® 8 Elite Mobile	4.126 ms	0 - 143 MB	NPU
DnCNN	ONNX	float	Snapdragon® 8 Elite Gen 5 Mobile	3.133 ms	0 - 141 MB	NPU
DnCNN	ONNX	float	Qualcomm® QCS9075	14.5 ms	0 - 46 MB	NPU
DnCNN	ONNX	float	Qualcomm® QCS8750	4.126 ms	0 - 143 MB	NPU
DnCNN	ONNX	float	Qualcomm® QCS7181	7.137 ms	180 - 180 MB	NPU
DnCNN	ONNX	w8a8	Snapdragon® X2 Elite	1.042 ms	213 - 213 MB	NPU
DnCNN	ONNX	w8a8	Snapdragon® X Elite	1.875 ms	149 - 149 MB	NPU
DnCNN	ONNX	w8a8	Snapdragon® 8 Gen 3 Mobile	1.337 ms	0 - 37 MB	NPU
DnCNN	ONNX	w8a8	Snapdragon® 8 Gen 1 Mobile	2.401 ms	0 - 44 MB	NPU
DnCNN	ONNX	w8a8	Qualcomm® QCS6490	7.86 ms	0 - 46 MB	NPU
DnCNN	ONNX	w8a8	Qualcomm® QCS8550 (Proxy)	1.797 ms	0 - 23 MB	NPU
DnCNN	ONNX	w8a8	Qualcomm® QCS8450	2.401 ms	0 - 44 MB	NPU
DnCNN	ONNX	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	0.76 ms	0 - 27 MB	NPU
DnCNN	ONNX	w8a8	Snapdragon® 7 Gen 4 Mobile	3.261 ms	0 - 143 MB	NPU
DnCNN	ONNX	w8a8	Qualcomm® QCM6690	40.034 ms	0 - 139 MB	NPU
DnCNN	ONNX	w8a8	Qualcomm® QCS9075	1.946 ms	0 - 46 MB	NPU
DnCNN	ONNX	w8a8	Snapdragon® 8 Elite Mobile	1.23 ms	0 - 30 MB	NPU
DnCNN	ONNX	w8a8	Qualcomm® QCS7790	3.261 ms	0 - 143 MB	NPU
DnCNN	ONNX	w8a8	Qualcomm® QCS8750	1.23 ms	0 - 30 MB	NPU
DnCNN	ONNX	w8a8	Qualcomm® QCS7181	1.875 ms	149 - 149 MB	NPU
DnCNN	QNN_DLC	float	Snapdragon® X2 Elite	4.176 ms	0 - 0 MB	NPU
DnCNN	QNN_DLC	float	Snapdragon® X Elite	7.206 ms	0 - 0 MB	NPU
DnCNN	QNN_DLC	float	Snapdragon® 8 Gen 3 Mobile	5.05 ms	0 - 177 MB	NPU
DnCNN	QNN_DLC	float	Snapdragon® 8 Gen 1 Mobile	13.731 ms	0 - 176 MB	NPU
DnCNN	QNN_DLC	float	Qualcomm® QCS8275	55.986 ms	0 - 141 MB	NPU
DnCNN	QNN_DLC	float	Qualcomm® QCS8550 (Proxy)	6.7 ms	0 - 2 MB	NPU
DnCNN	QNN_DLC	float	Qualcomm® QCS8450	13.731 ms	0 - 176 MB	NPU
DnCNN	QNN_DLC	float	Snapdragon® 8 Elite Mobile	4.035 ms	0 - 140 MB	NPU
DnCNN	QNN_DLC	float	Qualcomm® SA8295P	15.291 ms	0 - 139 MB	NPU
DnCNN	QNN_DLC	float	Snapdragon® 8 Elite Gen 5 Mobile	2.974 ms	0 - 145 MB	NPU
DnCNN	QNN_DLC	float	Qualcomm® SA7255P	55.986 ms	0 - 141 MB	NPU
DnCNN	QNN_DLC	float	Qualcomm® QCS9075	14.179 ms	2 - 4 MB	NPU
DnCNN	QNN_DLC	float	Qualcomm® QCS8750	4.035 ms	0 - 140 MB	NPU
DnCNN	QNN_DLC	float	Qualcomm® QCS7181	7.206 ms	0 - 0 MB	NPU
DnCNN	QNN_DLC	w8a8	Snapdragon® X2 Elite	1.172 ms	0 - 0 MB	NPU
DnCNN	QNN_DLC	w8a8	Snapdragon® X Elite	2.01 ms	1 - 1 MB	NPU
DnCNN	QNN_DLC	w8a8	Snapdragon® 8 Gen 3 Mobile	1.327 ms	0 - 46 MB	NPU
DnCNN	QNN_DLC	w8a8	Snapdragon® 8 Gen 1 Mobile	2.414 ms	0 - 53 MB	NPU
DnCNN	QNN_DLC	w8a8	Qualcomm® QCS6490	7.566 ms	0 - 2 MB	NPU
DnCNN	QNN_DLC	w8a8	Qualcomm® QCS8275	7.608 ms	0 - 28 MB	NPU
DnCNN	QNN_DLC	w8a8	Qualcomm® QCS8550 (Proxy)	1.792 ms	0 - 1 MB	NPU
DnCNN	QNN_DLC	w8a8	Qualcomm® QCS8450	2.414 ms	0 - 53 MB	NPU
DnCNN	QNN_DLC	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	0.753 ms	0 - 28 MB	NPU
DnCNN	QNN_DLC	w8a8	Snapdragon® 7 Gen 4 Mobile	3.255 ms	0 - 140 MB	NPU
DnCNN	QNN_DLC	w8a8	Qualcomm® QCM6690	38.856 ms	0 - 140 MB	NPU
DnCNN	QNN_DLC	w8a8	Qualcomm® QCS9075	1.928 ms	2 - 4 MB	NPU
DnCNN	QNN_DLC	w8a8	Qualcomm® SA7255P	7.608 ms	0 - 28 MB	NPU
DnCNN	QNN_DLC	w8a8	Snapdragon® 8 Elite Mobile	1.207 ms	0 - 30 MB	NPU
DnCNN	QNN_DLC	w8a8	Qualcomm® SA8295P	4.206 ms	0 - 26 MB	NPU
DnCNN	QNN_DLC	w8a8	Qualcomm® QCS7790	3.255 ms	0 - 140 MB	NPU
DnCNN	QNN_DLC	w8a8	Qualcomm® QCS8750	1.207 ms	0 - 30 MB	NPU
DnCNN	QNN_DLC	w8a8	Qualcomm® QCS7181	2.01 ms	1 - 1 MB	NPU
DnCNN	TFLITE	float	Snapdragon® 8 Gen 3 Mobile	5.199 ms	0 - 179 MB	NPU
DnCNN	TFLITE	float	Snapdragon® 8 Gen 1 Mobile	14.1 ms	1 - 180 MB	NPU
DnCNN	TFLITE	float	Qualcomm® QCS8275	56.398 ms	0 - 142 MB	NPU
DnCNN	TFLITE	float	Qualcomm® QCS8550 (Proxy)	6.946 ms	0 - 3 MB	NPU
DnCNN	TFLITE	float	Qualcomm® SA8775P	2195.166 ms	0 - 26 MB	GPU
DnCNN	TFLITE	float	Qualcomm® SA8650P	2195.166 ms	0 - 26 MB	GPU
DnCNN	TFLITE	float	Qualcomm® SA8255P	2195.166 ms	0 - 26 MB	GPU
DnCNN	TFLITE	float	Qualcomm® QCS8450	14.1 ms	1 - 180 MB	NPU
DnCNN	TFLITE	float	Snapdragon® 8 Elite Mobile	4.128 ms	0 - 143 MB	NPU
DnCNN	TFLITE	float	Qualcomm® SA8295P	15.612 ms	0 - 141 MB	NPU
DnCNN	TFLITE	float	Snapdragon® 8 Elite Gen 5 Mobile	3.097 ms	0 - 147 MB	NPU
DnCNN	TFLITE	float	Qualcomm® SA7255P	56.398 ms	0 - 142 MB	NPU
DnCNN	TFLITE	float	Qualcomm® QCS9075	14.225 ms	0 - 4 MB	NPU
DnCNN	TFLITE	float	Qualcomm® QCS8750	4.128 ms	0 - 143 MB	NPU
DnCNN	TFLITE	w8a8	Snapdragon® 8 Gen 3 Mobile	1.302 ms	0 - 46 MB	NPU
DnCNN	TFLITE	w8a8	Snapdragon® 8 Gen 1 Mobile	2.315 ms	0 - 51 MB	NPU
DnCNN	TFLITE	w8a8	Qualcomm® QCS6490	7.798 ms	0 - 3 MB	NPU
DnCNN	TFLITE	w8a8	Qualcomm® QCS8275	7.521 ms	0 - 29 MB	NPU
DnCNN	TFLITE	w8a8	Qualcomm® QCS8550 (Proxy)	1.721 ms	0 - 3 MB	NPU
DnCNN	TFLITE	w8a8	Qualcomm® SA8775P	2203.903 ms	1 - 26 MB	GPU
DnCNN	TFLITE	w8a8	Qualcomm® SA8650P	2203.903 ms	1 - 26 MB	GPU
DnCNN	TFLITE	w8a8	Qualcomm® SA8255P	2203.903 ms	1 - 26 MB	GPU
DnCNN	TFLITE	w8a8	Qualcomm® QCS8450	2.315 ms	0 - 51 MB	NPU
DnCNN	TFLITE	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	0.724 ms	0 - 29 MB	NPU
DnCNN	TFLITE	w8a8	Snapdragon® 7 Gen 4 Mobile	3.213 ms	0 - 141 MB	NPU
DnCNN	TFLITE	w8a8	Qualcomm® QCM6690	38.925 ms	0 - 140 MB	NPU
DnCNN	TFLITE	w8a8	Qualcomm® QCS9075	1.853 ms	0 - 3 MB	NPU
DnCNN	TFLITE	w8a8	Qualcomm® SA7255P	7.521 ms	0 - 29 MB	NPU
DnCNN	TFLITE	w8a8	Snapdragon® 8 Elite Mobile	1.175 ms	0 - 33 MB	NPU
DnCNN	TFLITE	w8a8	Qualcomm® SA8295P	4.179 ms	0 - 27 MB	NPU
DnCNN	TFLITE	w8a8	Qualcomm® QCS7790	3.213 ms	0 - 141 MB	NPU
DnCNN	TFLITE	w8a8	Qualcomm® QCS8750	1.175 ms	0 - 33 MB	NPU

License

The license for the original implementation of DnCNN 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.

Downloads last month: -; Downloads are not tracked for this model. How to track

Paper for qualcomm/DnCNN

Beyond a Gaussian Denoiser: Residual Learning of Deep CNN for Image Denoising

Paper • 1608.03981 • Published Aug 13, 2016