I am interested in Computer Vision, Real-time Graphics, Human-Computer Interaction, Video- and Image-Based Rendering.
My current research focuses on developing user-in-the-loop Computer Vision systems to help with video editing and understanding.
Scroll for Publications
Analysis and Synthesis of Interactive Video Sprites
2018, PhD Thesis awarded by UCL.
In this thesis, we explore how video, an extremely compelling medium that is traditionally consumed passively, can be transformed into interactive experiences and what is preventing content creators from using it for this purpose.
Film captures extremely rich and dynamic information but, due to the sheer amount of data and the drastic change in content appearance over time, it is problematic to work with. Content creators are willing to invest time and effort to design and capture video so why not for manipulating and interacting with it? We hypothesize that people can help and be helped by automatic video processing and synthesis algorithms when they are given the right tools.
Computer games are a very popular interactive media where players engage with dynamic content in compelling and intuitive ways. The first contribution of this thesis is an in-depth exploration of the modes of interaction that enable game-like video experiences. Through active discussions with game developers, we identify both how to assist content creators and how their creation can be dynamically interacted with by players. We present concepts, explore algorithms and design tools that together enable interactive video experiences.
Our findings concerning processing videos and interacting with filmed content come together in this thesis' second major contribution. We present a new medium of expression where video elements can be looped, merged and triggered interactively. Static-camera videos are converted into loopable sequences that can be controlled in real time in response to simple end-user requests. We present novel algorithms and interactive tools that enable our new medium of expression. Our human-in-the-loop system gives the user progressively more creative control over the video content as they invest more effort and artists help us evaluate it.
Monocular, static-camera videos are a good fit for looping algorithms but they have been limited to two-dimensional applications as pixels are reshuffled in space and time on the image plane. The final contribution of this thesis breaks through this barrier by allowing users to interact with filmed objects in a three-dimensional manner. Our novel object tracking algorithm extends existing 2D bounding box trackers with 3D information, such as a well-fitting bounding volume, which in turn enables a new breed of interactive video experiences. The filmed content becomes a three-dimensional playground as users are free to move the virtual camera or the tracked objects and see them from novel viewpoints.
Responsive Action-Based Video Synthesis
C Ilisescu, H. A. Kanaci, M. Romagnoli, N. Campbell, G.J. Brostow.
2017, ACM SIGCHI Conference on Human Factors in Computing Systems (CHI) 2017.
We propose technology to enable a new medium of expression, where video elements can be looped, merged, and triggered, interactively. Like audio, video is easy to sample from the real world but hard to segment into clean reusable elements. Reusing a video clip means non-linear editing and compositing with novel footage. The new context dictates how carefully a clip must be prepared, so our end-to-end approach enables previewing and easy iteration.
We convert static-camera videos into loopable sequences, synthesizing them in response to simple end-user requests. This is hard because a) users want essentially semantic-level control over the synthesized video content, and b) automatic loop-finding is brittle and leaves users limited opportunity to work through problems. We propose a human-in-the-loop system where adding effort gives the user progressively more creative control. Artists help us evaluate how our trigger interfaces can be used for authoring of videos and video-performances.