|
Activity detection and recognition We are interested in detecting, recognizing and classifying human motion and human activities from the least amount of information (frames) as possible, as well as detecting rare events, which is an inherently ill-defined problem. AVSS 2010 [PDF] (Graph matching) ICPR 2010a [PDF] (Bags of pairwise features) ICPR 2010b [PDF] (BG subtraction) Technical report LIRIS-2010 [PDF] (Shape sequences and deep belief nets) ICANN 2010 [PDF] (Long-short term memory networks)
|
||||||
|
Image segmentation and restoration We deal with image segmentation and restoration taking into account as much a priori information on the image as possibly available. Probabilistic graphical models (Markov random field, Bayesian networks etc.) are particularly well suited to this task and naturally combine with very efficient global optimization methods (graph cuts, belief propagation etc.). IEEE-T-PAMI 2010 [PDF] Neurocomputing 2010 [PDF] Ranked 5th of 43 in the ICDAR 2009 document image binarisation contest Pattern Analysis and Applications 2003[PDF] ICPR 2002[PDF] |
||||||
|
Mesh Analysis We are working on the segmention of 3D triangular meshes and on remeshing, i.e. optimization of the triangle shapes of an existing mesh while keeping the best possible surface approximation. Our methods are based on the optimization of gobal discrete energy functions, for instance through graph cut techniques. The Visual Computer (to appear) [PDF] GRAPP 2011 Eurographics-W3DOR 2008 [PDF] TR LIRIS 2009 [PDF] |
||||||
|
Image retargeting, content aware zooming Maximize the information visible in a limited area (screen space). Graphics Interface 2010 [PDF] |
||||||
|
Object detection and image indexing Our activities in this domain involve detection and recognition of 2D and 3D objects in images and videos, pose estimation, detection of object classes, as well as image classification. CBMI 2009a[PDF] CBMI 2009b[PDF] Pattern Analysis and Applications 2003[PDF] IJDAR 2005 |
||||||
|
Evaluation of object detection algorithms We have proposed a new performance graphs for the evalaluation of object detection algorithms, which illustrate performance intuitively by graphs which present object level precision and recall depending on constraints on detection quality. In order to compare different detection algorithms, a representative single performance value is computed from the graphs. The evaluation method can be applied to different types of object detection algorithms. It has been tested on different text detection algorithms, among which are the participants of the ICDAR 2003 text detection competition. IJDAR 2006[PDF] [Software] This method has been used to evaluate the participants in two scientific competitions: ICDAR 2003 Robust Reading Competition [REF] Image Eval 2007 text detection competition [REF] |
||||||
