Researchers have developed a computer algorithm that can have an underwater image and automatically readjust the colors to compensate for the distorting impact of water on light.
Researchers Derya Akkaynak and Tali Trebitz began work on the technology called “Sea-Thru” over 3 years back. Akkaynak Sea-thru’s mission is enabling huge, artificial powered analysis of underwater images.
The algorithm effectively changes underwater pictures to make them look as they had been clicked in daylight which makes them easier for AI applications to analyze.
“On underwater pictures, AI methods typically perform inconsistently or poorly, because water degrades pictures way too seriously for automated analysis,” she said.
This usually implies marine scientists have to personally pore through pictures to analyze what pictures they have captured, which may prove to be a huge task.
“It is not unusual they come back again to the lab with thousands to thousands from a one-time dive. Just a small portion of those pictures ever get examined, simply because on average, it requires a man expert 2 hours to determine and count fish in a video clip which is one hour long,” she stated.
Trebitz, who established the University of Haifa’s marine-imaging lab, included that the manner light is distorted in water may be difficult to recognize when you are diving because the human mind adjusts to it.
“We are both avid divers, so when you plunge the human brain shockingly compensates for the water effects. So there are cases where you dive and encounter a colorful and rich tremendously atmosphere, but when you take a look at the photographs you acquired after the plunge they look blue and dull, and you understand that there’s more there than what’s really apparent in the images,” she stated.
While at the University of Haifa in Israel in 2015, Akkaynak realized observed that scientists had been using color correction equations created to be used in the atmosphere.
“Once I found that I formulated a (more) physically accurate equation especially for the ocean, and this is called the Akkaynak Treibitz model is the true breakthrough which led to the Sea-thru algorithm,” she stated.
Akkaynak tried the algorithm by placing color charts next to underwater objects. She then ran the image through Sea thru and was in a position to determine if the chart had been changed back to its proper colors. The finished Sea Thru solution does not need color charts to run.
Sea-thru might have a broad range of uses based on Akkaynak “It requires away the importance to carry artificial lights, which means less gear and expense to have for several photographers”, she added.
“I will be used in Photoshop, incorporated into consumer cameras, and diving masks,” she included. However, Akkaynak Sea-thru might enable a new AI-driven trend of marine data analysis, although Sea-thru itself does not make use of any AI.
“When these pictures (e.g. surveys of reefs, seafloor, fish, fish stocks) are pre-processed with Sea-thru, scientists will then have the ability to make use of existing effective computer vision and machine learning techniques to have the ability to count, segment, identify, and classify other objects and animals in them. This’s the reason I see Sea-thru as the beginning of the AI boom of marine science,” she stated.
With modern science understanding much more about the surface area of the moon as compared to what it does on the subject of the depths of this ocean, Sea-thru might be the key element to some remarkable future scientific discoveries.