[ad_1]
Can artificial intelligence and a 3D printer help fight the biggest threat in the art world?
A team from the MIT Computer Science and Artificial Intelligence Laboratory (CSAIL) thinks so.
A new system aimed at designing reproductions of paintings uses 3D printing and deep learning to "authentically recreate favorite paints", regardless of the lighting conditions and their placement.
"If you only reproduce the color of a painting as it appears in the gallery, its appearance might be different at home," said study co-author Changil. Kim, postdoctoral researcher at CSAIL.
Thus, instead of using the four traditional fixed inks (cyan, magenta, yellow, black) present in 2D printers, the researchers applied a special technique called "color matching."
The team formed a deep learning model to predict the optimal stack of different inks (via MIT CSAIL).
The process involves a 3D printer and 10 different transparent inks piled in very thin layers, similar to wafers and chocolate from a Kit-Kat bar.
By combining this method with a decades-old semitone approach (which creates a gradient effect through dots), they have been able to better capture the "nuances of colors".
Based on experimental reproductions of various oil paintings created by a collaborating artist, the team found that RePaint was four times more accurate in creating accurate color shades than physical models at the cutting edge. technology.
"Our system works in all lighting conditions, which shows a much greater color reproduction capability than almost any previous work," Kim said.
However, there is one disadvantage: the facsimiles of CSAIL are about the size of a business card. After all, 3D printing is not expensive.
However, in the future, they expect advanced commercial printers to produce larger paints, paving the way for a more efficient system.
The question remains: which inks should be used for which paints?
As with many tasks nowadays, humans have transferred the burden of selection to an in-depth learning model, which can predict the optimal stack of shades in each area of the canvas.
The creators of RePaint plan to use it for the reconstruction of works of art for the home, the protection of originals in museums and the creation of copies of historical objects.
The researchers found that RePaint is more than four times more accurate than the most modern physical models for creating the exact color shades of different works of art (via MIT CSAIL).
The program has a long way to go before it can start producing duplicate Starry Night.
For starters, he can not completely reproduce certain colors – like cobalt blue – because of a limited ink library, which mechanical engineer Mike Foshey hopes to develop soon.
And, as you can see in the images and videos above, the imitations of RePaint are clearly missing: the texture. The team will continue to improve the details, hoping to create special effects such as glossy and matte finishes.
"The value of works of art has grown rapidly in recent years. So it's becoming increasingly common to lock them up in warehouses, away from the public eye, "Foshey said. "We are building technology to reverse this trend and create accurate, inexpensive reproductions that everyone can enjoy."
A full report was posted this week by MIT CSAIL.
More coverage on Geek.com:
[ad_2]
Source link
