CS Events

PhD Defense Reconstructing 3D Humans and Animals from Visual Data
Export Event Save iCal Send to Google Calendar Send to Yahoo Calendar Send to Outlook Live Send to Microsoft Outlook Save iCal Preserve formatting in description (only supported in some calendar applications) Close Friday, July 19, 2024, 10:00am - 11:30am

Speaker: Anastasios Stathopoulos

Location : CoRE 301

Committee: 

Prof. Dimitris Metaxas (chair)

Prof. Karl Stratos

Prof. Konstantinos Michmizos

Prof. Sharon Xiaolei Huang (external)

Event Type: PhD Defense

Abstract: This dissertation explores the problem of reconstructing humans and animals from visual data. Given the large number of applications that require a 3D model as input like augmented and virtual reality (AR/VR), film production, and forensics, the ability to obtain accurate 3D models from images is more relevant than ever. Recent advancements in 3D human mesh recovery have significantly enhanced our ability to reconstruct humans in 3D with remarkable precision. Despite these successes, current 3D perception systems still lag behind their 2D counterparts in terms of robustness. In this talk, I will first outline our latest efforts to enhance the accuracy of 3D human recovery, with the help of automatically detected image observations (e.g., 2D keypoint detections), and address the limitations of existing work for this task. Then, I will present our work on 3D animal reconstruction. I will discuss the challenges associated with this task, particularly the absence of ground truth 3D data and the scarcity of 2D annotations, and propose weaker forms of supervision that are effective in overcoming this limitation.

: 

Contact  Professor Dimitris Metaxas (Chair)

Zoom Log In: https://rutgers.zoom.us/j/97713374085?pwd=Edv18FwuZJqp0myYaxynX8HLzKMvLB.1