Research Topics

[Topic] Few-shot Learning and Model Adaptation

Introduction Despite recent success of deep neural networks, it remains challenging to efficiently learn new concepts from limited training data. To address this problem, our group has been working on several novel few-shot learning and model adaptation strategies, focusing on the structural representation of input data.

[Topic] Holistic Video Understanding

Introduction We address the problem of multiclass semantic video segmentation, aiming to integrate object reasoning, scene layout estimation with pixel labeling. To this end, we have developed an object-augmented structured model in spatio-temporal domain, which captures long-range dependency between superpixels, and imposes consistency between object and superpixel labels.

[Topic] Object and Scene Segmentation

Introduction We address the problem of joint detection and segmentation of multiple object instances in an image, which is a key step towards scene understanding. In particular, we aim to reduce the annotation cost and integrate prior knowledge into data-driven methods.

[Topic] Visual Relationship and Vision-Language Representation

Introduction Reasoning about the relationships between objects is a crucial task for holistic scene understanding. Beyond existing works of recognition and detection, relationships between objects also constitute rich semantic information about the scene.