Research
Face Recognition
We are conducting research on facial recognition technology, which includes face detection, 1:1 verification, and 1:N identification. Additionally, we have submitted our developed methods to the Face Recognition Vendor Test (FRVT) 1:1 Verification benchmark maintained by the US National Institute of Standards and Technology (NIST).
We propose S3PNet, a 3D CNN architecture that learns shared 2D triplanar features viewed from three orthogonal planes. It has fewer parameters compared to linearly stacked 3D CNNs, enabling it to learn 3D representations without redundancy.
This research presents a new framework for creating retinal photomontages using deep learning-based object detection and vessel segmentation for registration and blending. The framework involves two steps: rigid and non-rigid registration.
Vessel Segmentation & Artery/Vein Classification
We propose a new framework for retinal vessel segmentation by registering and segmenting corresponding FA images, which consists of three subprocesses: registration and vessel extraction of FA frames, multimodal registration of aggregated FA vessels to the fundus image, and post-processing for refining the vessel mask.
Vessel Graph Network for Retinal Vessel
We present a method to localize and classify masses from BUS images by training a CNN on a relatively small dataset with strong annotations and a large dataset with weak annotations in a hybrid manner.
Joint Weakly- and Semi-Supervised Deep Learning
We present a novel CNN architecture, the vessel graph network (VGN), that jointly exploits the global structure of vessel shape together with local appearances.
