Computer Vision Misc Reading Group

Year 2003    (in chronological order)

Date	Presenter	Description
1/29/2003 Wednesday 12:00 noon NSH 4201	Shyjan Mahamud	For our next meeting, I will be talking about applications of loopy belief propagation in computer vision. First, good tutorials on belief propagation (BP) can be found: here. In particular the tech. report. I will then review 3 applications of BP that I have found: "Recovering Shading and Reflectance from a Single Image", M. Tappen, W.T. Freeman, E.H. Adelson, NIPS 2002. Paper here. "Finding Deformable Shapes Using Loopy Belief Propagation", J.M. Coughlan, S.J. Ferreira, ECCV 2002. Paper here. "Enforcing Integrability for Surface Reconstruction Algorithms Using Belief Propagation in Graphical Models", N. Petrovic, I. Cohen, B. Frey, R. Koetter, T.S. Huang, CVPR 2001. Paper here. I may not be able to go into much detail on all of these applications in one talk as I will be focusing more on reviewing the theory first. However, if there is enough interest, we can talk about these papers in more detail in another meeting.
2/5/2003 Wednesday 12:00 noon NSH 4201	Martial Hebert	I will talk about some topics on closest point search and approximate NN recently making its way into C.V. applications. Related Papers: P. Indyk and R. Motwani, "Approximate nearest neighbor - towards removing the curse of dimensionality", In Proceedings of the 30th Symposium on Theory of Computing, 1998. Paper. Recognizing 3D Objects from Range Data. Andrea Frome, Thomas Bulow, and Jitendra Malik. Abstract. (Submitted to CVPR 2003.) Additional info: Buhler. Bioinformatics 17(5) 419-428, 2001 Shyjan's Thesis http://dias.cti.gr/~ytheod/research/indexing/ http://www.cs.sunysb.edu/~algorith/major_section/1.6.shtml http://www.cs.washington.edu/homes/jbuhler/ http://theory.lcs.mit.edu/~indyk/webdb.ps
2/12/2003 Wednesday 2:00 pm NSH 3001	Daniel Huber	Here is the paper that we will be reading: Torr and Zisserman, MLESAC: A new robust estimator with application to estimating image geometry, CVIU 78:1, April 2000, p. 138-56. Paper here. Zhengyou Zhang at INRIA has a nice tutorial on parameter estimation in HTML and Postscript.
2/19/2003 Wednesday 2:00 pm NSH 3001	Pragyana Mishra	Here is the abstract of the talk I will be giving at this week's reading group. Since it is regarding my recent work, I do not have a paper for your perusal. A Simple Approach to Alignment without Correspondence This talk will present a new and generalized framework to to align a model with respect to an image. The approach does not explicitly require the nature of imaging or model registration process, illumination and reflectance conditions, surface properties of the scene, determining occlusions and correspondence of features between data from sensors. The method is based on a new formulation of minimizing a distance measure relating the statistics of two randomly sampled data sets, in our case, a range model and an intensity image. Sensor data have regions best suitable for computing a statistic while parts of it are cluttered or unresolved. The framework exploits the General Crofton Theorem in integral geometry to combine statistics from disparate local regions of the data. This new approach can accommodate data from different sensors and can be used to align data of same or varied modalities.
2/26/2003 Wednesday 2:00 pm NSH 3001	Fernando De La Torre	Next week I will present a couple of papers (the second one just if I have enough time). Both papers are related to constructing generative models from image sequences. Christopher K.I. Williams and Michalis K. Titsias. Learning About Multiple Objects in Images: Factorial Learning without Factorial Search. Paper here. Y. Z. Wang, S.C. Zhu, "A Generative Method for Textured Motion: Analysis and Synthesis", Proc. of European Conf. on Computer Vision (ECCV), Copenhagen, June, 2002. Paper here. Further material here.
3/5/2003		No Meeting.
3/12/2003		No Meeting.
3/19/2003 Wednesday 2:00 pm NSH 3305	Caroline Pantofaru	Dorin Comaniciu. An Alogorithm for Data-Driven Bandwidth Selection. PAMI, vol. 2, no. 2, pp. 281-288, February 2003. Paper here.
3/26/2003 Wednesday 2:00 pm NSH 3001	Ranjith Unnikrishnan	"Robust Computer Vision through Kernel Density Estimation", Haifeng Chen and Peter Meer, ECCV 02. Paper here.
4/2/2003 Wednesday 2:00 pm NSH 3001	Derek Hoiem	FloatBoost is similar to Adaboost except that instead of merely adding weak classifiers to the strong classifier, weak classifiers may be removed from the classifier as well. The intent is to overcome some of the greediness of the Adaboost method and to find a smaller set of weak classifiers that achieves equal or greater performance. The FloatBoost paper introduces FloatBoost, a more flexible feature set than Viola-Jones, and a hierarchical detector for multi-view detection of faces: S.Z. Li, L. Zhu, Z.Q. Zhang, A. Blake, H.J. Zhang, H. Shum. "Statistical Learning of Multi-View Face Detection". In Proceedings of The 7th European Conference on Computer Vision. Copenhagen, Denmark. May, 2002. Paper here and here.
4/9/2003 Wednesday 2:00 pm NSH 3001	Chuck Rosenberg	I am going to discuss the object detection work of David Lowe using local scale-invariant features. We've mentioned this work a number of times but we've never actually read the papers in the group. Martial has already presented much of the background material in his talk about a year ago on 7/10/2002. The basic paper is: Object recognition from local scale-invariant features. David G. Lowe, International Conference on Computer Vision, Corfu, Greece (September 1999), pp. 1150-1157. Paper here. This paper extends the concept to better handle variations in 3D pose: Local feature view clustering for 3D object recognition, David G. Lowe, IEEE Conference on Computer Vision and Pattern Recognition, Kauai, Hawaii (December 2001), pp. 682-688. Paper here.
4/16/2003 Wednesday 2:00 pm NSH 3001	Owen Carmichael	For the next misc reading group I'll be talking about the super-resolution of image sequences, i.e. inferring a high-resolution image from a sequence of overlapping low-resolution images. (This is in contrast to the single-image super-resolution problem, where the goal is to infer a high-resolution image from a single low resolution image plus prior knowledge or training data). The reference is: Michael E Tipping and Christopher M Bishop. Bayesian Image Super-resolution. Proceedings NIPS 2002. Paper here. I will discuss how this relates to earlier work along the same line by Peter Cheeseman, Blake and Zisserman, and others, time permitting.
4/23/2003 Wednesday 2:00 pm NSH 3001	Anuj Kapuria	I have selected a paper, from PAMI, August 2002 (Vol. 24, No. 8), pp. 1075-1090. The title is: Discovering Useful Concept Prototypes for Classification Based on Filtering and Abstraction. Wai Lam, Chi-Kin Keung and Danyu Liu. Paper here.
4/30/03 Wednesday 2:00 pm NSH 3001	Sanjiv Kumar	I will be talking about the work of Torralba, Murphy, Billman and Rubin on the use of context for object detection/recognition: http://www.ai.mit.edu/~murphyk/Papers/AImemo03.pdf
5/7/03 Wednesday 2:00 pm NSH 1305	Edward H. Adelson	Special RI Seminar On Seeing Stuff: The Perception of Materials by Humans and Machines
5/14/03 Wednesday 2:00 pm NSH 3001	David Tolliver	Kernel Principal Angles for Classification Machines with Applications to Image Sequence Interpretation. Lior Wolf and Amnon Shashua. CVPR 2003. Paper here.
5/21/03 Wednesday 2:00 pm NSH 3001	Goksel Dedeoglu	This Wednesday I would like to talk about the following paper: Storkey A.J., "Dynamic structure super-resolution", NIPS 2002. Paper here. Comparison with other techniques is available here.
5/28/2003		No Meeting.
6/4/2003 Wednesday 2:00 pm NSH 3001	Bart Nabbe	"An Efficient Solution to the Five-Point Relative Pose Problem", David Nister, Sarnoff Corporation, CVPR 2003.
6/11/2003 Wednesday 2:00 pm NSH 3001	Raghu Rao	I shall be talking about a variant of NN classifier. The two papers are: Stan Z Li, J.Lu, 'Face Recognition Using the Nearest Feature Line Method', IEEE Trans on Neural Networks, vol.10, no.2, pp.439-443, Mar.1999. Paper here. and J.T. Chien, C-C, Wu, 'Discriminant Waveletfaces and nearest feature classifiers for face recognition', IEEE Trans. PAMI, 24(12):1644-1649. 2002. Paper here. In the first paper, they introduce, Nearest_Feature_Line (NFL) classifier a variant of Nearest Neighbor Classifier (NN), in which any two feature points of the same class (person's face) are generalized by a feature line (FL) passing through the two points. They found that this performs better than simple NN classifier. In the second paper, they extend this idea to Nearest_feature_plane (NFP) and Nearest_feature_space (NFS). NFP considers sets of three features in the class, while NFS considers subspaces of any dimension. Turns out NFP/NFS do better than the others (NFL and NN).
6/25/2003 Wednesday 2:00 pm NSH 3002	Everyone	CVPR 2003 Conference Sweep The goal of this meeting is to compile a list of papers from CVPR that we would like to read and understand in more detail in coming meetings and have a short discussion about the overall trends at CVPR.
7/2/2003 Wednesday 2:00 pm NSH 3001	Cristian Dima	The workshop is called "Multiple Classifier Systems", and this year's edition of the program is here. I will go over the papers I have found interesting and if (my) time permits I will also try to go over the key points in the invited talk by Friedman.
7/9/2003 Wednesday 2:00 pm NSH 3001	Shyjan Mahamud	For the next group meeting, I will be giving a talk on Helmholtz Stereopsis for the uncalibrated case: "Towards a Stratification of Helmholtz Stereopsis", T. Zickler, P. Belhumeur, D. Kriegman, CVPR 2003. Paper here. I will also review related work at the same conference: "Surface Reconstruction via Helmholtz Reciprocity with a Single Image Pair",P. Tu, P. Mendonca, CVPR 2003.
7/16/2003 Wednesday 2:00 pm NSH 3001	Daniel Huber	For next week's reading group, I'll discuss: Object Class Recognition by Unsupervised Scale-Invariant Learning by Fergus, Perona, and Zisserman. CVPR 2003. Paper here. Hopefully, it will give us some ideas for our paper on scale-invariant beta-boosting. ;)
7/23/2003 Wednesday 2:00 pm NSH 3001	Fernando De La Torre	Next week I will try to understand (no guarantees are given) the following paper: Optimal Linear Representations of Images for Object Recognition X. Liu, A. Srivastava, K. Gallivan (Florida State University) (CVPR 2003) Paper here. In case I have some extra time, you some interest and it is before 3 p.m I will also try to go through a related paper: Statistical Search for Hierarchical Linear Optimal Representations of Images Q. Zhang, X. Liue, A. Srivastava. (IEEE CVPR 2003 Workshop on Statistical Analysis in Computer Vision) Paper here.
7/30/2003 Wednesday 2:00 pm NSH 3001	Pragyana Mishra	I will present the paper on Vector-Valued Image Regularization with PDE's: A Common Framework for Different Applications. by D. Tschumperle and R. Deriche. Paper here. The paper got the 'best student paper award' at CVPR03.
8/6/2003		No meeting because of the VASC Retreat.
8/13/2003 Wednesday 2:00 pm NSH 3001	Ranjith Unnikrishnan	I'll be introducing the paper "Practical Non-parametric Density Estimation on a Transformation Group for Vision", E. Miller, C. Chefd'hotel (CVPR '03) (paper here) that talks about estimating densities of linear shape deformations for the purpose of OCR. For motivation / background on the classifier model they use for some of their results, read "Learning from One Example Through Shared Densities on Transforms.", E.Miller, N.Matsakis, P.Viola, (CVPR '00), (paper here). I'll try and cover both papers, time and sanity permitting.
8/20/2003 Wednesday 2:00 pm NSH 3001	Caroline Pantofaru	I'm going to discuss: Edge Detection with Embedded Confidence. Peter Meer, Bogdan Georgescu. PAMI, vol. 23, no. 12, December 2001. Paper here.
8/27/2003 Wednesday 2:00 pm NSH 3001	Derek Hoiem	I'll present "Kullback-Leibler Boosting" (CVPR 2003) by Liu and Shum from Microsoft on Wednesday (Sanjiv said he will be covering different papers). If there is time, I may also cover "An efficient approach to learning inhomogeneous Gibbs model" (CVPR 2003) by Liu, Chen, and Shum. The first paper describes an approach to feature generation and selection and classification using the KL-divergence metric and boosting. The second paper describes a method for fast parameter learning of Inhomogeneous Gibbs model in high dimensional spaces. The first paper also describes an application in face detection, while the second describes an application in the all-important domain of caricature generation. The papers are available online: Paper 1 (local copy) and Paper 2 (local copy).
9/3/2003		No Meeting
9/10/2003 Wednesday 2:00 pm NSH 3001	Chuck Rosenberg	I'll be presenting: "A Bayesian Approach to Unsupervised One-shot Learning of Object Categories" by Li Fei-Fei, Rob Fergus, Pietro Perona. Pages 1134-1141. ICCV 2003. Paper here.
9/17/2003 Wednesday 2:00 pm NSH 4201	David Tolliver	I'll present Stella Yu's paper on K-Way Cuts: Multiclass Spectral Clustering. Stella Yu and Jianbo Shi. Paper here.
9/24/2003 Wednesday 2:00 pm NSH 3001	Sanjiv Kumar	I will speak on nonparametric belief propagation to do inference over arbitrarily connected graphs with continuous latent variables: Nonparametric Belief Propagation. E. Sudderth, A. Ihler, W. Freeman, and A. Willsky. CVPR, June 2003. Paper here.
10/1/2003		No Meeting.

Last modified: Mon Apr 29 2004