Machine learning generates 3D model from 2D pictures

Researchers from the McKelvey School of Engineering at Washington University in St. Louis have developed a machine learning algorithm that can create a continuous 3D model of cells from a partial set of 2D images that were taken using the same standard microscopy tools found in many labs today.

Their findings were published Sept. 16 in the journal Nature Machine Intelligence.

"We train the model on the set of digital images to obtain a continuous representation," said Ulugbek Kamilov, assistant professor of electrical and systems engineering and of computer science and engineering. "Now, I can show it any way I want. I can zoom in smoothly and there is no pixelation."

The key to this work was the use of a neural field network, a particular kind of machine learning system that learns a mapping from spatial coordinates to the corresponding physical quantities. When the training is complete, researchers can point to any coordinate and the model can provide the image value at that location.

A particular strength of neural field networks is that they do not need to be trained on copious amounts of similar data. Instead, as long as there is a sufficient number of 2D images of the sample, the network can represent it in its entirety, inside and out.

The image used to train the network is just like any other microscopy image. In essence, a cell is lit from below; the light travels through it and is captured on the other side, creating an image.

"Because I have some views of the cell, I can use those images to train the model," Kamilov said. This is done by feeding the model information about a point in the sample where the image captured some of the internal structure of the cell.

Then the network takes its best shot at recreating that structure. If the output is wrong, the network is tweaked. If it's correct, that pathway is reinforced. Once the predictions match real-world measurements, the network is ready to fill in parts of the cell that were not captured by the original 2D images.

The model now contains information of a full, continuous representation of the cell -- there's no need to save a data-heavy image file because it can always be recreated by the neural field network.

And, Kamilov said, not only is the model an easy-to-store, true representation of the cell, but also, in many ways, it's more useful than the real thing.

Read more at Science Daily

New tool makes students better at detecting fake imagery and videos

 Researchers at Uppsala University have developed a digital self-test that trains users to assess news items, images and videos presented on social media. The self-test has also been evaluated in a scientific study, which confirmed the researchers' hypothesis that the tool genuinely improved the students' ability to apply critical thinking to digital sources.

The new tool and the scientific review of it are part of the News Evaluator project to investigate new methods of enhancing young people's capacity for critical awareness of digital sources, a key component of digital literacy.

"As research leader in the project, I'm surprised how complicated it is to develop this type of tool against misleading information -- one that's usable on a large scale. Obviously, critically assessing digital sources is complicated. We've been working on various designs and tests, with major experiments in school settings, for years. Now we've finally got a tool that evidently works. The effect is clearly positive and now we launch the self-test on our News Evaluator website http://www.newsevaluator.com, so that all anyone can test themselves for free," says Thomas Nygren, associate professor at Uppsala University.

The tool is structured in a way that allows students to work with it, online, on their own. They get to see news articles in a social-media format, with pictures or videos, and the task is to determine how credible they are. Is there really wood pulp in Parmesan cheese, for instance?

"The aim is for the students to get better at uncovering what isn't true, but also improve their understanding of what may be true even if it seems unlikely at first," Nygren says.

As user support, the tool contains guidance. Students can follow how a professional would have gone about investigating the authenticity of the statements or images -- by opening a new window and doing a separate search alongside the test, or doing a reverse image search, for example. The students are encouraged to learn "lateral reading" (verifying what you read by double checking news). After solving the tasks, the students get feedback on their performance.

When the tool was tested with just over 200 students' help, it proved to have had a beneficial effect on their ability to assess sources critically. Students who had received guidance and feedback from the tool showed distinctly better results than those who had not been given this support. The tool also turned out to provide better results in terms of the above-mentioned ability than other, comparable initiatives that require teacher participation and more time.

Read more at Science Daily