The Journal Club meets every 2 weeks on Wednesday afternoons.
Presentations are expected to be on the board, no slides should be prepared. Think of it as an elaborate poster presentation, but without the poster, and that the audience will have seen the paper you are presenting. Exceptions are if you are presenting someone's paper which has associated slides or videos etc. already available, in which case they may be used. In extreme circumstances, you may wish to use the JC as a dry run of a conference talk you are about to give. A PDF of the paper should be submitted to the JC webmaster at least one week in advance of the talk.
The next talk is coloured in green.
|2011||April 13||Jack Greenhalgh||
J. Shotton, A. Fitzgibbon, M. Cook, T. Sharp, M. Finocchio, R. Moore, A. Kipman, A. Blake – Real-Time Human Pose Recognition in Parts from Single Depth Images|
Quantum Room (MVB 3.44), 2pm to 3pm
We propose a new method to quickly and accurately predict 3D positions of body joints from a single depth image, using no temporal information. We take an object recognition approach, designing an intermediate body parts representation that maps the difficult pose estimation problem into a simpler per-pixel classification problem. Our large and highly varied training dataset allows the classifier to estimate body parts invariant to pose, body shape, clothing, etc. Finally we generate confidence-scored 3D proposals of several body joints by reprojecting the classification result and finding local modes.
The system runs at 200 frames per second on consumer hardware. Our evaluation shows high accuracy on both synthetic and real test sets, and investigates the effect of several training parameters. We achieve state of the art accuracy in our comparison with related work and demonstrate improved generalization over exact whole-skeleton nearest neighbour matching.
|January 19||Osian Haines||
R. Subbarao, Y. Genc and P. Meer, Nonlinear Mean Shift for Robust Pose Estimation.|
Abstract: We propose a new robust estimator for camera pose estimation based on a recently developed nonlinear mean shift algorithm. This allows us to treat pose estimation as a clustering problem in the presence of outliers. We compare our method to RANSAC, which is the standard robust estimator for computer vision problems. We also show that under fairly general assumptions our method is provably better than RANSAC. Synthetic and real examples to support our claims are provided.
Quantum Meeting Room, from 3pm to 4pm.
|2010||August 18||John Bartholomew||
Hinton, G. E., Osindero, S. and Teh, Y A fast learning algorithm for deep belief nets. [Extra info: YouTube video.]|
Meeting Room (3.36) from 2 to 3pm
|June 09||Irwandi Hipiny||
Shai Avidan and Ariel Shamir SIGGRAPH 2007 paper Seam Carving for Content-Aware Image Resizing. [Extra info: YouTube video and website with more information] |
Quantum Meeting room (3.44) from 12 to 1pm
|May 26||Dima Damen||
Image Segmentation by Branch-and-Mincut. Efficient global optimization techniques such as graph cut exist for energies corresponding to binary image segmentation from low-level single pixel information. Introducing a high-level prior such as a shape prior into the segmentation process results in an energy function that is much harder to optimize. The main contribution of the paper is to combine graph-cut with a branch-and-bound algorithm search algorithm for adding shape priors to object segmentation. This is achieved through the derivation of lower bounds on the energies. Being computable via graph cut, these bounds are used to prune branches within a branch-and-bound search. Results show superior performance over the previous Grabcut segmentation method. |
Quantum Meeting room (3.44) from 2 to 3pm
|May 12||Sudeep Sundaram||Balanced Graph Matching. Graph matching is a fundamental problem in Computer Vision and Machine Learning. We present two contributions. First, we give a new spectral relaxation technique for approximate solutions to matching problems, that naturally incorporates one-to-one or one-to-many constraints within the relaxation scheme. The second is a normalization procedure for existing graph matching scoring functions that can dramatically improve the matching accuracy. It is based on a reinterpretation of the graph matching compatibility matrix as a bipartite graph on edges for which we seek a bistochastic normalization. We evaluate our two contributions on a comprehensive test set of random graph matching problems, as well as on image correspondence problem. Our normalization procedure can be used to improve the performance of many existing graph matching algorithms, including spectral matching, graduated assignment and semidefinite programming. More info here.|
|March 17||Jose Martinez-Carranza||Real-Time Learning of Accurate Patch Rectification|
|March 03||Osian Haines||Depth estimation from image structure. In the absence of cues for absolute depth measurements as binocular disparity, motion, or defocus, the absolute distance between the observer and a scene cannot be measured. The interpretation of shading, edges, and junctions may provide a 3D model of the scene but it will not provide information about the actual "scale" of the space. One possible source of information for absolute depth estimation is the image size of known objects. However, object recognition, under unconstrained conditions, remains difficult and unreliable for current computational approaches. Here, we propose a source of information for absolute depth estimation based on the whole scene structure that does not rely on specific objects. We demonstrate that, by recognizing the properties of the structures present in the image, we can infer the scale of the scene and, therefore, its absolute mean depth. We illustrate the interest in computing the mean depth of the scene with application to scene recognition and object detection.|
|February 17||Sudeep Sundaram||Boundary Extraction in Natural Images Using Ultrametric Contour Maps. This paper presents a low-level system for boundary extraction and segmentation of natural images and the evaluation of its performance. We study the problem in the framework of hierarchical classification, where the geometric structure of an image can be represented by an ultrametric contour map, the soft boundary image associated to a family of nested segmentations. We define generic ultrametric distances by integrating local contour cues along the regions boundaries and combining this information with region attributes. Then, we evaluate quantitatively our results with respect to ground-truth segmentation data, proving that our system outperforms significantly two widely used hierarchical segmentation techniques, as well as the state of the art in local edge detection. Further info.|
|February 03||Mosalam Ebrahimi||Robust Face Recognition via Sparse Representation. Recently compressive sensing (CS), a new mathematical framework for sampling and reconstructing sparse signals, has been used for some computer vision tasks with very promising results. This seminal paper uses CS for face recognition, and their method outperforms the state-of-the-art strikingly. More info about this work.|
|January 20||Adeline Paiement||This talk will be an introduction to the use of level sets for image segmentation, with a look at Statistical shape influence in geodesic active contours which gives an example of a segmentation method using level sets and a shape model.|
|2009||November 25||Xiaosong (Sean) Wang||The prize winning BMVC 2009 paper Stochastic Image Denoising by Estrada, Fleet and Jepson.|
|November 11||John McGonigle||An introduction to empirical mode decomposition and a look at Image analysis by bidimensional empirical mode decomposition. [Extra info: a good overview and the original paper]|
|October 28||Pished Bunnun||Looking at the Georg Klein and David Murray ISMAR '07 paper Parallel Tracking and Mapping for Small AR Workspaces.|
Computer Vision Group
Dept of Computer Science,
University of Bristol
For more information about our work or opportunities to join or visit the group, email vision at cs.bris.ac.uk