Project 51: Multimodal Event Detection Under Unparalleled Situations
Contact Information
Assoc. Prof. Mengyue Wu
Email: mengyuewu@sjtu.edu.cn
Project Description and Objectives
Audio-based event detection is challenging due to data imbalance and various durations of audio events. This project investigates how to introduce video information for multi-modal event detection and improve the robustness of event detection. However, with unparalleled data, it is critical to explore the fusion method according to the scenario and select corresponding information according to the modal confidence at different times.
Eligibility Requirements
Interested students should have basic knowledge of machine learning.
Main Tasks
Development of a paradigm to fuse audio and video for robust event detection.
Completion of a research report.
Website