Afshin Dehghan

PhD student (since Fall 2011)
Computer Vision Lab
Supervised by Mubarak Shah
University of Central Florida

Email: adehghan8@cs.ucf.edu (Remove 8)
Office: Harris Corporation Engineering Center (HEC), Room 254
Research interests: Vision, Machine Learning

 

Publications


GMCP-Tracker: Global Multi-object Tracking Using Generalized Minimum Clique Graphs
Amir Roshan Zamir, Afshin Dehghan and Mubarak Shah
In proceedings of European Conference on Computer Vision 2012 (ECCV'12)
[PDF] [Project Page] [Data]

Abstract: Data association is an essential component of any human tracking system. The majority of current methods, such as bipartite matching, incorporate a limited-temporal-locality of the sequence into the data association problem, which makes them inherently prone to ID-switches and difficulties caused by long-term occlusion, cluttered background, and crowded scenes. We purpose a global approach to data association which incorporates both motion and appearance in a global manner. Unlike limited-temporal-locality methods which incorporate 2 or few frames into the data association problem, we incorporate the whole temporal span and solve the data association problem for one object at a time, while implicitly incorporating the rest of the objects. In order to achieve this, we utilize Generalized Minimum Clique Graphs to solve the optimization problem of our proposed data association method. Our proposed method yields a better formulated approach to data association supported by our superior results. Experiments show the proposed method makes significant improvements in tracking in the diverse sequences of Town Center, TUD-crossing, TUDStadtmitte , PETS2009, and a new sequence called Parking Lot compared to the state of the art methods.

 

Keynote: Automatic Detection and Tracking of Pedestrians in Videos with Various Crowd Densities
Afshin Dehghan, Haroon Idrees, Amir Roshan Zamir and Mubarak Shah
In Proceedings of PED, June 2012
[PDF] [Project Page] [Data]

Abstract: Manual analysis of pedestrians and crowds is often impractical for massive datasets of surveillance videos. Automatic tracking of humans is one of the essential abilities for computerized analysis of such videos. In this keynote paper, we present two state of the art methods for automatic pedestrian tracking in videos with low and high crowd density. For videos with low density, first we detect each person using a part-based human detector. Then, we employ a global data association method based on Generalized Graphs for tracking each individual in the whole video. In videos with high crowd-density, we track in-dividuals using a scene structured force model and crowd flow modeling. Addi-tionally, we present an alternative approach which utilizes contextual infor-mation without the need to learn the structure of the scene. Performed evalua-tions show the presented methods outperform the currently available algorithms on several benchmarks.

 

Part-based Multiple-Person Tracking with Partial Occlusion Handling
Guang Shu, Afshin Dehghan, Omar Oreifej, Emily Hand, Mubarak Shah
In proceedings of IEEE Conference on Computer Vision and Pattern Recognition (CVPR'12)
[PDF] [Project Page] [Data]

Abstract: Single camera-based multiple person tracking is a very challenging problem. The performance of state-of-the-art methods is often hindered by difficulties such as occlusion and changes in appearance which frequently occurin surveillance videos. In this paper, we address such problems by proposing a robust part-based tracking-bydetection framework. Human detection using part models has become quite popular, yet its extension in tracking has not been fully explored. Our approach learns partbased person-specific SVM classifiers which capture the articulations of the human bodies in dynamically changing appearance and background. The confidence from the person classifier is employed along with motion and size cues to construct an affinity matrix which is used to associate the observations with the tracking through a greedy bipartite approach. Our method handles partial occlusions in both the detection and the tracking stages. In the detection stage, we obtain a probabilistic representation of the parts and accordingly select the subset of parts which maximizes the probability of detection, which significantly improves the detection performance in crowded scenes. In the tracking stage, we dynamically handle occlusions by distributing the score of the learned person classifier among its corresponding parts, which allows us to detect and predict partial occlusions, and prevent the performance of the classifiers from being degraded. Extensive experiments using the proposed method on several challenging sequences demonstrate state-of-the-art performance in multiple-people tracking.

 

A Multi-Agent Architecture for Tracking User Interactions in Browser-based Games
A.A. Bagherzadeh, S. Rezvankhah, S. Farahi, K. Khalvati, P. Mousavi, A. Dehghan, B. Ghaderi, L. Kashani, H. Moradi
In proceedings of IEEE International Conference On Digital Game And Intelligent Toy Enhanced Learning (DIGITEL12)
[PDF] [Project Page] [Data]

Abstract: This paper presents a tracking system developed for browser-based games to track user interactions remotely. To eliminate the restriction of accessing local file system, the client side of the system runs on a standalone web server. The tracking agent writes the user interactions with a game on the local file system, and the log-transfer agent sends the results to the server side of the system as soon as an internet connection is available. This system is used in a research to investigate the effects of playing certain games on visual working memory of children under the age 7. This tracking system has provided the games with better user-friendliness by removing browser's permission prompts in JavaScript and Flash-based games. Furthermore, allowing remote data collection from all the gamers provides vast data for future research.