Muralikrishna Sridhar

Post Doctoral Research Fellow
E-mail: krishna@comp.leeds.ac.uk

Current Research

My current research focusses on extending and refining the framework developed during my PhD, for unsupervised learning of event and object classes from video.

The goal of this unsupervised framework is to simultaneously learn the partitioning of a video scene into events (particular interactions), while learning the corresponding event classes (classes of similar interactions).

Interactions are represented using a relational graph based representation that we refer to as interaction graphs. They have an equivalent logical representation.

Furthermore, functional object classes are learned from the learned event classes. Objects beloning to the same functional object class tend to have similar functions.

This framework has been demonstrated on complex video scenes such as an airport apron. An example of a learned event class is unloading, which represents characteristic interactions between objects such as loaders, trolleys and planes. An example of a learned functional object class is trolleys, since they perform a similar function with respect to the unloading event class.

Previous Research

I completed my PhD at the University of Leeds in 2010. My thesis focussed on pointing a camera at a complex scene and learning the event classes and object classes by modelling interactions between objects. Before comming to leeds, I was the owner of the Shape Recognition Project at Hewlett Packard Labs, India (2005-2007). I worked on a combination of generative and discriminative kernel based classifiers for the recognition of complex handwritten characters and gestures. I also spent a year doing my MSc by Research in Machine Learning at the University of London (2004-2005).My MSc research focussed on unsupervised learning of word classes from a text corpus. This technique provided a starting point during my PhD, for learning functional object classes from Video.

Publications

	Muralikrishna Sridhar, Anthony Cohn, Hogg David C, Benchmarking Qualitative Spatial Calculi for Video Activity Analysis, Proc. IJCAI Workshop (2011). Video of Spatio-Temporal Graphs >
	Muralikrishna Sridhar, Anthony Cohn, Hogg David C, From Video to RCC8: exploiting a Distance Based Semantics to Stabilise the Interpretation of Mereotopological Relations, Proc. COSIT (2011). BibTeX Best Paper Award for Cosit 2011. Video of RCC8-HMM >
	Muralikrishna Sridhar, Anthony Cohn, Hogg David C, Unsupervised Learning of Event Classes from Video, Proc. AAAI (2010). BibTeX Video of Learned Event >
	Muralikrishna Sridhar, Anthony Cohn, Hogg David C, Discovering an Event Taxonomy from Video using Qualitative Spatio-temporal Graphs, Proc. ECAI (2010).
	Muralikrishna Sridhar, Anthony Cohn, Hogg David C, Relational Graph Mining for Learning Events from Video, Proc. STAIRS (2010).
	Muralikrishna Sridhar, Anthony Cohn, Hogg David C, Learning Functional Object-Categories from a Relational Spatio-Temporal Representation, Proc. ECAI (2008). Sample Video of Toy Kitchen
	Muralikrishna Sridhar, Dinesh M and Mehul P, Active-DTW: A Generative Classifier that combines Elastic Matching with Active Shape Modeling for Online Handwriting Recognition, Proc. ICFHR (2006).
	Dinesh Mandalapu and Muralikrishna Sridhar A Feature based on Encoding the Relative Position of a Point in the Character for Online Handwritten Character Recognition, Proc. ICDAR (2007).
	Muralikrishna Sridhar, Deepu V, Dinesh M and Sriganesh M, A Generic Approach for Handwritten Shape Recognition, HP Labs Technical Report, Sept (2005).
	Muralikrishna Sridhar, Unsupervised Learning of Event and Object Classes from Video, PhD thesis, University Of Leeds, UK, 2010
	Muralikrishna Sridhar, Word Similarity Modelling with Slot Phrases, MSc thesis, University Of London, UK, 2004

Software

REDVINE: RElational Description of VIdeo SceNEs (comming soon !) Video Of Spatial Relations