UNIVERSITY PARK, Pa. — Baris Kandemir, a doctoral candidate in Penn State's College of Information Sciences and Technology, is spending his summer interning at DeepMap, a startup company in Palo Alto, California, that makes high-definition recognition maps for autonomous cars.
Creating these maps is a crucial step in the development of the emerging technology of autonomous vehicles, but it’s no easy task. It requires advanced equipment and precision measurements to accurately map the environment and avoid catastrophe.
“If you try to use a normal camera, the images [don’t have] 3D information surrounding the car,” Kandemir said. “They need accuracy and they need to decide what to do with the accurate information.”
A native of Izmir, Turkey, Kandemir got his start with a bachelor’s degree in electrical engineering from Bogazici University in his home country. Upon entering his doctoral program at IST with his adviser James Wang, Kandemir began working in image processing and studied how computers can be developed to recognize different visual elements.
“Baris has been working on two highly interdisciplinary topics in his doctoral dissertation research,” said Wang. One of Kandemir's projects, according to Wang, is an interdisciplinary collaboration with the Department of Biology and the Department of Agricultural and Biological Engineering.
“[This project] studies stomatal guard cell walls using an integrated molecular, dynamic imaging and mechanical modeling approach,” said Wang. Baris has been developing 3-D microscopic image analysis methods.
Wang explained that Kandemir’s research, and subsequently his application of the research, can have implications to many fields of technology, such as robotics, affective computing, multimedia, and agricultural production.
In his internship at DeepMap, Kandemir is able to apply this work in 3D image analysis to help further the technology behind autonomous vehicles. He explained that car companies all have their own self-driving departments that focus on hardware; however, they all require the technology DeepMap is developing in order for them to work.
“The cars heavily rely on the map system to know where they are,” Kandemir said. While cameras in these vehicles initially depended on the car camera, he explained, it isn’t the most efficient way to obtain the information. “They aren’t working well,” said Kandemir, “so they started looking into different ways to supply that information.”
The maps are created using a collection of data from process imaging, a gathering method known as "deep learning."
“It’s composed of neutral networks, and for this model, you try to find the correct things in the image,” Kandemir said.
This summer, he is tasked with improving training data for the technology, which Kandemir believes will be used by different car companies in the future.
“We’ll be the base supplier for all these automobile companies. In the end, it will make better self-driving cars,” he said, noting that this technology will help to prevent accidents and vehicular-related deaths.
Kandemir said that, in addition to preparing him for this internship, his time in IST has given him valuable skills to apply in the workforce, and he expressed gratitude for the role his Penn State professors play in his education.
“[The professors] are pushing everyone to try to be better," he said. "Overcoming failures and trying again and again — these are qualities I learned in IST.”
He added that his internship is the perfect example of how working for a startup, even as an intern, can be a positive and fulfilling experience.
“They’re working hard and they are also really friendly people,” he said of his colleagues at DeepMap. “It’s very warm, as it’s a small company, and it’s growing faster. It’s great to be with a company like this.”
Kandemir explained that this technology is still in the early phases, but sees its prospects. “I think we still have some way to go,” he said, “but it has the potential to change the whole scenery [of self-driving cars.]”