A Breakthrough in Segmenting Images Presented at the ICLR

In April, PhD student Bin Wang of the Bagci lab presented at the International Conference on Learning Representations (ICLR): “Order-aware Interactive Segmentation to Explicitly Integrate Missing Relative Depth Information into 2D Interactive Segmentation.”  

Wang is a third-year PhD student in electrical and computer engineering, supervised by Dr. Ulas Bagci. He earned his B.Eng. degree in computer science from ShanghaiTech University. 

Wang’s algorithm enables people to interactively segment 2D images. Previous algorithms would often capture parts of an image that were peripheral to the targeted object, requiring considerable work to filter out what wasn’t wanted. Wang’s work utilizes information not previously explored to find the relative distance between objects in an image and isolate the targeted body.    

The algorithm also creates a higher resolution image that shows many more details than current algorithms. Such a breakthrough has global applications in the medical field and in any industry that needs to isolate still images or images in videos.  

Achievements such as Wang’s not only increase the Radiology Department’s visibility but elevate our department’s academic standing. The ICLR is one of the most prestigious AI conferences globally, ranking among top journals such as Nature Medicine, JAMA, Cell, and the Proceedings of the National Academy of Sciences. It is a premier learning conference that requires evidence of multiple experiments that demonstrate the effectiveness of an algorithm.