Skip navigation links

Jan. 16, 2018

Yuan Liang and Biyi Fang: Improving satellite maps

Jan. 17, 2018

Biyi Fang and Yuan Liang are doctoral students studying electrical and computer engineering. Fang's research focus is deep learning optimization and implementation on mobile devices, while Liang's focus is the modeling and optimization of wireless networks and communications. Both students are part of the Institute for Cyber-Enabled Research and use the High Performance Computing Center for their work.

MSU students Biyi Fang and Yuan Liang are working toward solving a problem regarding the pixel-wise segmentation on satellite maps. Basically, objects are interlaced on maps and there are multiple objects in small areas.

They aim to separate objects and differentiate them at the same time. They do this by isolating the pixel and giving a label to the specific type of terrain it belongs to, whether that be buildings, crops, waterways, etc. Fang and Liang’s main focus is to be able to recognize objects captured on a satellite map, and then label these types of objects along with where they are located.

Both Fang and Liang believe that the direction of the future regarding electrical and computer engineering lies in deep learning image recognition. This, along with their general interest, is how they chose their research topic.

Due to the development of Graphics Processing Unit, amongst many other things, the algorithms in deep learning give people the chance to do something that was impossible in the past. One of the things that made this possible is recognizing the images of different objects.

When it comes to long term goals, Fang and Liang hope that their research will help people in the future. They think that if the work they’ve done attracts more people, their method would be implemented into terrain recognition, which would render into military usage. It’s very complicated and expensive for people to recognize objects one by one and picture by picture, so the label work that Fang and Liang are doing is important to the field.

“We could not even have begun this project, let alone finish it, without the help of iCER and HPCC,” says Fang. “Our research requires a deep learning network that’s used for a long time, along with a great amount of computational power. The HPCC provides us, along with other students, a simple and economic way to use that computational power that we otherwise wouldn’t.”

Both students say that the freedom they have and the resources that are so conveniently located have been momentous in their research so far.

“I have had the opportunity to meet different people and hear about what they’re doing,” says Liang.  “It allows me to expand my knowledge on different views”.

Reused with modifications and permission from the Institute for Cyber-Enabled Research