Research

The Embedded Deep Learning and Visual Analysis Laboratory relies on several national and Shanghai key laboratories of the School of Information Science and Engineering of Fudan University, and focuses on the research and development of various deep learning algorithms for mobile, edge or chipset ASIC-level platforms. These algorithms are applied but not limited to: object detection, object classification, scene analysis, 3D reconstruction, and object tracking applications, and play an important role in various scenarios such as smart homes, smart cities, medical AI, ADAS and security monitoring that are closely related to society and people’s everyday life.

Currently, the lab is focusing on several research fields including but not limited to:

  1. Lightweight deep learning study, multimodal deep learning research. PAGCP Vote2CapDETR-MLD

  2. Embedded deep learning: The main research is to design a small-scale deep learning network with lightweight, low complexity and low memory usage under limited computing and memory conditions such as mobile, edge or ASIC custom chips. research2

  3. Visual computing and content analysis of images, videos, etc.

Deep learning algorithm for semantic classification of images and videos, target detection and segmentation; analysis or reconstruction of video scenes; tracking of key targets and other common visual applications. Typical application scenarios include: target monitoring based on pedestrians or faces , traffic management and planning based on vehicle detection and attribute analysis research3

We also have collaborators from both academia and industries: collaborators