DnCNN3 | Luxonis

Our new model ZOO works with DepthAI V3. Find out more in our documentation.

1+ Likes

Model Details

Model Description

DnCNN3 (denoising CNN) is a convolutional neural network designed for image denoising. It demonstrates high effectiveness in various denoising tasks as, for example, Gaussian denoising.

Developed by: Kai Zhang et al.
Shared by:
Model type: Computer Vision
License:
Resources for more information:

Photo by Matthew Jackson from

Training Details

Training Data

91 images from and 200 training images from the .

Testing Details

Metrics

PSNR (Peak Signal-to-Noise Ratio) and SSIM (Structural Similarity Index) metrics were performed for various conditions to measure the denoizing quality. Here we report the results for Gaussian denoising task evaluated on dataset at three different noise levels (σ). See the for more information.

Metric	σ=15	σ=25	σ=50
PSNR	31.46	29.02	26.10
SSIM	0.8826	0.8190	0.7076

Technical Specifications

Input/Output Details

Input:
- Name: image
  - Info: NCHW grayscale image
Output:
- Name: enhanced_image
  - Info: Denoised version of the input image

Model Architecture

Modifified VGG architecture (see the for more information).

Throughput

Model variant: dncnn3:640x480

• Input shape: [1, 1, 480, 640] • Output shape: [1, 1, 480, 640]

• Params (M): 0.666 • GFLOPs: 204.945

Platform	Precision	Throughput (infs/sec)	Power Consumption (W)
RVC4	INT8	83.26	6.82

Model variant: dncnn3:320x240

• Input shape: [1, 1, 240, 320] • Output shape: [1, 1, 240, 320]

• Params (M): 0.666 • GFLOPs: 51.236

Platform	Precision	Throughput (infs/sec)	Power Consumption (W)
RVC2	FP16	3.05	N/A
RVC4	INT8	362.31	8.04

* Benchmarked with , using 2 threads (and the DSP runtime in balanced mode for RVC4).

* Parameters and FLOPs are obtained from the package.

Quantization

RVC4 version of the model was quantized using a HubAI General dataset.

Utilization

Models converted for RVC Platforms can be used for inference on OAK devices. DepthAI pipelines are used to define the information flow linking the device, inference model, and the output parser (as defined in model head(s)). Below, we present the most crucial utilization steps for the particular model. Please consult the docs for more information.

Install DAIv3 and depthai-nodes libraries:

pip install depthai
pip install depthai-nodes

Define model:

model_description = dai.NNModelDescription(
    "luxonis/dncnn3:320x240"
)

nn = pipeline.create(ParsingNeuralNetwork).build(
    <CameraNode>, model_description
)

Inspect model head(s):

ImageOutputParser that outputs message (enhanced image).

Get parsed output(s):

while pipeline.isRuning():
    parser_output: dai.ImgFrame = parser_output_queue.get()

Example

You can quickly run the model using our script. It automatically downloads the model, creates a DepthAI pipeline, runs the inference, and displays the results using our DepthAI visualizer tool. To try it out, run:

python3 main.py \
    --model luxonis/dncnn3:320x240

Image denoising model.
License	MIT Commercial use
Downloads	511
Tasks	Denoising
Model Types	ONNX

Model Variants

Name	Version	Available For	Created At	Deploy
		RVC4	7 months ago
		RVC2, RVC4	7 months ago