ICME is one of the flagship conferences in multimedia. In this competition, participants are asked to predict whether a user will finish and like a specific short video along with its multi-modal features. My solution is ranked $3^{rd}$ out of 1028 teams from all over the world. See competition LeaderBoard.
WSDM is one of the premier conferences on web-inspired research involving search and data mining. One of the task in WSDM Cup 2019 is to detect the fake news. Given the title of a fake news article A and the title of a coming news article B, participants are asked to classify B into one of the three categories. My method is ranked 6/8 (public/private) among 94 teams from all over the world. See Kaggle competition.
The project aims at designing search algorithms to retrieval personal photos given text query. The work has been published in CLEF 2018 with title Visual Concept Selection with Textual Knowledge for Understanding Activities of Daily Living and Life Moment Retrieval. For the details of the project, please refer to the official pages.
In this task, we have to plot bounding boxes for each disease finding in a single chest X-ray without goundtruth (X, Y, width, height) in training set. The task description is detailed in PDF1 and PDF2. Our work flow could be summarized as 
Cross-domain image retrieval is a challenging problem due to the data variations between the real-world images and advertisement images. In this work, we consider four state-of-the-art deep learning based model to extract the high-level features combining with four feature pooling strategies. Different from previous works, we further investigate the possibility of integrating the classical feature descriptors. A dataset containing half a million images of beauty and care products (Perfect-500k) is utilized for our experiments. The experimental results prove that our proposed hybrid framework can improve the mAP@7 between 3% and 10% in contrast with retrieval methods only utilizing deep features. Product pictures from different domains are show in (a) and (b) below. 
Taiwan is considered as a small island comparing to her population. Besides, with high population density in the city area such as in Taipei city, New Taipei city, it is quite difficult to find an available parking space in those places. Therefore, we want to propose a mobile application that can provide parking lot information to customer with value-added services such as the availability of the parking lot and information about how long the customer needs to wait if the parking lot is full. With our proposed mobile application, the customer can plan their schedule in advance which can increase their satisfaction in using the facility. Moreover, with the forecasting approximate waiting time, the customer will be prepared for waiting which can help lower their frustration. Additionally, the parking lot company can better manage their parking lots via the user log in the mobile application. 