Yale-NUS Assistant Professor of Science Robby Tan was interviewed by Lianhe Zaobao on his latest research in computer vision. Dr Tan developed a new deep-learning algorithm which is able to remove rain streaks in videos that are captured. This algorithm helps to improve other technologies in autonomous vehicles or driverless cars which rely on video images, combatting image quality issues caused by bad weather such as rain.
Dr Tan used more than 10,000 images to train the computer system to distinguish rain streak patterns, in order to remove them effectively. While there are still some limitations, such as a darker output due to the reflection of light by rain as well as the occasional removal of background objects mistakenly identified as rain, Dr Tan shared that the research team is working on ways to optimise the technology and resolve such issues. The article noted that the “rain wiper” improvement can also be used in other technologies such as surveillance cameras. The article also noted that Dr Tan spent about eight months developing the algorithm, and presented the findings in Hawaii earlier in July at the Conference on Computer Vision and Pattern Recognition.