Using GANs For Underwater Color Images
Written by David Conrad   
Sunday, 12 January 2020

Neural networks, in particular GANs (Generative Adversarial Networks) have found yet another application. This time it's underwater vision for the exploration of seabed resources, fishing and underwater archaeology.

The new research comes from Harbin Engineering University in China and starts from the premise that the main reason for poor underwater image quality is the scattering and attenuation of light. The scattering results in haze effect while the attenuation of light leads to color cast.

Taking a machine-learning approach to the problem the researchers used an unsupervised generative adversarial network (GAN) to create realistic underwater images, with color distortion and haze effect from in-air images and depth map pairs.


The GAN was trained on a corpus of labeled scenes containing 3,733 images chiefly of scallops, sea cucumbers, sea urchins, and other such organisms living within indoor marine farms, plus corresponding depth maps, The team also sourced open data sets, including NY Depth which comprises thousands of underwater photographs in total.

They then utilized U-Net, which has been trained using synthetic underwater dataset, for color restoration and dehazing. According to the paper, the model:

directly reconstructs underwater clear images using end-to-end autoencoder networks while maintaining scene content structural similarity.


The results obtained by this approach was compared with existing methods qualitatively and quantitatively. The paper states:

Experimental results obtained by the proposed model demonstrate well performance on open real-world underwater datasets, and the processing speed can reach up to 125FPS running on one NVIDIA 1060 GPU.

As well as documenting the research in a paper published on arxiv the source code and sample datasets have been made publicly available on GitHub.


More Information

UWGAN: Underwater GAN for Real-world Underwater Color Restoration and Dehazing (pdf)

UWGAN on GitHub

Related Articles

GANPaint: Using AI For Art

GANalyze - What Makes Pictures Memorable?

Sea-Thru Removes Water From Undersea Images

SoFi - A Robot Fish Contributes to Scientific Research 

ReefScouts - Swarm Water Robotics 

To be informed about new articles on I Programmer, sign up for our weekly newsletter, subscribe to the RSS feed and follow us on Twitter, Facebook or Linkedin.



Microsoft's Generative AI for Beginners

A free, self-paced online course about Generative AI is on offer from Microsoft's Azure Cloud Advocates. It's a 12-lesson curriculum targeted at complete novices to LLMs.

AWS Lambda Adopts Java 21

AWS Lambda functions can now use all the new and useful language features as well as performance improvements introduced in Java 21 as part of the Amazon Corretto JDK implementation.

More News




or email your comment to:


Last Updated ( Sunday, 12 January 2020 )