Research
My research focuses on imbalanced image recognition. In imbalanced image recognition, data are not balanced or curated because they are gathered in large volumes and annotation of all object instances is intractable. In this scenario, the data may be long-tailed, which means that there may be objects that appear frequently in the dataset while there may be other that are unique and do not appear frequently. Under this realistic scenario, models trained with imbalanced data may have good performance in identifying the frequent objects of the dataset but they may fail to identify rare objects. My reseach aims to develop methods that tackle the long tailed recognition problem, so that the models can have better performance in detecting and recognising all the objects in the dataset.
I also serve as a reviewer for CVPR, ICCV and ECCV since 2022.
Example of Object Detection