Salman Anwar Khokhar

 

Graduate Research Assistant

 

Center For Research in Computer Vision, UCF

 

 

 

 

Publications:

1 - Similarity invariant classification of events by KL divergence minimization, Salman Khokhar, Imran Saleemi, Mubarak Shah, ICCV 2011. [code]

2 - Multi agent event recognition by preservation of spatio-temporal relationships between probabilistic models, Salman Khokhar, Imran Saleemi, Mubarak Shah, IVC 2013.

3 - Turn prediction at generalized intersections , Bo Tang, Salman Khokhar, Rakesh Gupta, IV 2015

4 - Tracking when the camera looks away , Khurram Soomro, Salman Khokhar, ICCVW 2015

 

Patents:

1 – (Pending) Turn predictions, USPTO Serial: 14/801904, Filed July 17, 2015. R. Gupta, S. Khokhar, B. Tang

 

 

 

 

 

 

 

 

 

Short Bio:

 

I am a PhD candidate at UCF working under the supervision of Prof. Mubarak Shah.

                       

Resume

Selected Research Projects:

Scene modeling using motion patterns: PerSeas

 

 

In this project we use track or optical flow data to extract motion patterns from a scene which serve as descriptors of the activity in the scene. These patterns are obtained by a hierarchical linking of clusters of flow or tracking data. They can then be used to detect anomalous behavior in traffic flow.

Project page

 

 

Similarity invariant classification of events by KL divergence minimization: PerSeas

 

In this project we aim to model events found in traffic scenes using motion patterns obtained from a hierarchical clustering of optical flow. Two events are then compared by warping them in such a way that their KL divergence is minimized. We accomplish this using an iterative warping approach. After this a dissimilarity score may be computed using KL divergence.

Related publication , Project page, Presentation

 

 

Multi-agent event recognition by preservation of spatiotemporal relationships between probabilistic models

 

In this project we aim to model multi-agent activities using a graphical model that encodes agent behaviors as graph nodes and their inter-relationships as graph edges.

Related publication , Project page, Presentation [Slides, Video]

 

 

                        Contact:

·         Email: s”mylastname” ‘at’ eecs . ucf . edu

·         Ph: 407 962 5895

·         Physical address: HEC 254, UCF, Central FL Blvd, Orlando, FL, 32816, USA.