Research Group
Dr John Collomosse is a reader (assoc. prof.) in computer vision within CVSSP and a visiting professor at the Adobe Creative Intelligence Lab. Collomosse's Ph.D. (Bath, 2004) studied the use of computer vision to parse salient structure from scenes, and to use these structures as the basis for creating synthetic artwork (e.g. paintings and cartoons). He continues to actively research these and related Graphics topics, but in the past years has also focused on the inverse problem of mapping drawings (e.g. sketches) to photo-realistic footage. This has led to several funded projects exploring the use of sketch (and other visual query modalities) for intuitive search of large visual media collections.
Dr Andrew Gilbert is a post-doctoral research fellow within the group, working on the InnovateUK funded Total Capture and REFRAME projects in collaboration with BBC, The Imaginarium, Numerion and Imagineer Systems. His Ph.D. (Surrey, 2008) proposed new algorithms for scalable and adaptable tracking of objects. On his current projects Andy is exploring multi-viewpoint video techniques for in-painting as part of a real-time system for live performance capture and on-set previz of virtual characters. He co-supervises the PhD of Mat Trumble on deep learning techniques for human pose estimation in multiple viewpoint video. This work is aligned with complementary work within CVSSP on 3D Reconstruction via the Visual Media Lab (VML) and studio facilities.
Yifan Yang is a post-doctoral research fellow working on the EPSRC funded 'TAPESTRY' project on social media behaviour and identity analysis. His work on TAPESTRY explores robust machine learning algorithms for pattern of life analysis in social media content `in the wild'. His Ph.D. thesis (Telecom ParisTech, 2016) was on machine learning and signal processing.
Tu Bui is a final year Ph.D. student working on large scale visual search and classification problems for multi-million record image databases. His research has so far contributed scalable representations for large-scale sketch based image retrieval (SBIR) and for extending SBIR to additional modalities beyond shape e.g. colour. His ongoing work explores deep learning representations for sketch based image retrieval.
Mat Trumble is a final year Ph.D. student working on human pose estimation from multi-viewpoint video. His work contributes to CVSSP's 4D performance capture capability and explores the automated rigging of deformable surface models captured via this technology with skeletons so facilitating the use of traditional animation pipelines with 4D data. His ongoing work focuses on deep learning for human pose estimation in multiple viewpoint video sequences.
Jongdae (JD) Kim is a final year Ph.D. student working on the application of transfer learning techniques to large scale object recognition, and incrementally learnable classifiers. He is working with the semantic ImageNet ontology to tackle the problem of large-scale visual concept detection in video. Recent work also addressed the task of video summarisation and thumbnails for large-scale video collection browsing and was presented orally at Eurographics 2015.
Alumni
Previous members of our research group:
- Dr Stuart James (Visual narratives for sketch based image and video search)
- Dr Tinghuai Wang (Video segmentation and Artistic Rendering w. HP Labs)
- Mr Charles Gray (Scalable Video Concept Detection and Annotation (MPhil)
- Dr Rui Hu (Sketch Based Image Retrieval)
- Dr Qizhi Yu (Kinect based character animation / motion retargetting)
- Dr Yu (Jade) Qian (Video Visual Search)
- Dr Graham McNeill (Video Visual Search)
- Dr Yi-Zhe Song (Artistic Rendering and Cellphone based 3D reconstruction)