X

Name	Version	Available For	Created At	Deploy
		RVC2, RVC4	About 1 year ago

Model Details

Model Description

Zero-Reference Deep Curve Estimation (Zero-DCE) is a deep neural network for enhancing low-light images using light-enhancement curve estimation. We implement here the full (Zero-DCE) version of the model.

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

Training Details

Training Data

900 custom training dataset of low-light images, 9000 unlabeled dataset low-light images, and 4800 SICE dataset images. See the for more information.

Testing Details

Metrics

Different test metrics were calculated on and datasets. See the for more information.

Metric	Value	Dataset
Signal-to-Noise Ratio (PSNR,dB)	6.57	SICE
Structural Similarity (SSIM)	0.59	SICE
Mean Absolute Error (MAE)	98.78	SICE
full-precision accuracy	16.24	LOL

Technical Specifications

Input/Output Details

Input:
- Name: image
  - Info: NCHW BGR low-light image
Output:
- Name: enhanced_image
  - Info: Brightness and contrast enhanced version of the input image

Model Architecture

Custom architecture (see the for more information on model architecture).

Throughput

Model variant: zero-dce:600x400

• Input shape: [1, 3, 400, 600] • Output shape: [1, 3, 400, 600]

• Params (M): 0.079 • GFLOPs: 19.745

Platform	Precision	Throughput (infs/sec)	Power Consumption (W)
RVC2	FP16	4.74	N/A
RVC4	INT8	57.86	3.30

* 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/zero-dce:600x400"
)

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

Inspect model head(s):

ImageOutputParser that outputs message (light-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/zero-dce:600x400

Low-light image enhancement model.
License	MIT Commercial use
Downloads	491
Tasks	Low Light Enhancement
Model Types	ONNX