Andrew Gilbert
  • Home
  • About & Contact
  • Publications
  • Work Areas
    • Stylised In-painting
    • 3D Pose estimation
    • iGroup
    • Action Recognition
    • Multi touchless and Hand Gestures
    • Surveillance
  • TotalCapture Dataset

What I do...

Creative Visual Media production is a large and growing industry, and as a result have a number of research challenges within it. Multiple viewpoint video is becoming commonplace in film and music video production, where shots are captured using multiple, wide-spaced, cameras.  Within this i have two research questions
  1. is it possible to remove an actor in a motion-capture suit to be later replaced with an animated virtual character. Illustrated by the music video from the Chemical Brothers Wide Open,  and Alicia Vikander in Ex Machina,
  2. Can I perform real time human pose estimation of actors, however, freeing the actor from having to wear a cumbersome and environment limiting motion capture suit and therefore allowing the pose estimation to occur in free environments and in costume. For current solutions to this check out recent state of the art performance on Human3.6M dataset published at BMVC 2017 and 3DV 2017

Previously...

What I have researched before:
​Every day, users struggle with the ever increasing quantity of data available to them. Trying to find “that” photo they took on holiday last year, the video on YouTube they were watching at their friends last week, or the news article someone mentioned at work. I am working on developing approaches that can automatically understand what is happening in huge amounts of very diverse types of media including, photos, documents, audio and videos, without needing to tell the computer beforehand what it is looking at.
Picture

Picture
Picture
Picture

Updates

  • ECCV'18 accepted on 
  • CVPR'18 Poster on Disentangling Structure and Aesthetics for Style-aware Image Completion
  • 3DV 2017 oral on 3D human Pose estimation using IMU sensors
  • BMVC 2017 Oral on 3D human Pose estimation 
  • New 3D Pose estimation dataset: "TotalCapture" for multi-view video, IMU and vicon groundtruth ~1.9Million images
  • Accepted Image and Video Mining through Online Learning journal to CVIU, 
  • Accepted Hand Pose estimation journal to CVIU, based on a discriminative classifier and model regression​
  • Home
  • About & Contact
  • Publications
  • Work Areas
    • Stylised In-painting
    • 3D Pose estimation
    • iGroup
    • Action Recognition
    • Multi touchless and Hand Gestures
    • Surveillance
  • TotalCapture Dataset