:::::: 2009 "Humies" Awards for Human-Competitive Results :::::: "Evolutionary Learning of Local Descriptor Operators for Object Recognition" ------------------------------------------------------------------------------------------------------ 1. The complete title of one (or more) paper(s) published in the open literature describing the work that the author claims describes a human-competitive result: Perez C.B., Olague G. "Learning Invariant Region Descriptor Operators with Genetic Programming and the F-measure". International Conference on Pattern Recognition. Pages 1-4. ISBN: 978-1-4244-2174-9. December 8-11, 2008. Perez C.B., Olague G. "Evolutionary Learning of Local Descriptor Operators for Object Recognition". To appear in Genetic and Evolutionary Computation Conference. July 8-12, 2009. ------------------------------------------------------------------------------------------------------ 2. The name, complete physical mailing address, e-mail address, and phone number of EACH author of EACH paper: Cynthia B. Perez EvoVision Laboratory Centro de Investigacion Cientifica y de Educacion Superior de Ensenada (CICESE) Km. 105 Tijuana-Ensenada Ensenada B.C., Mexico cbperez@cicese.mx Gustavo Olague EvoVision Laboratory Centro de Investigacion Cientifica y de Educacion Superior de Ensenada (CICESE). Km. 105 Tijuana-Ensenada Ensenada B.C., Mexico olague@cicese.mx ------------------------------------------------------------------------------------------------------ 3. The name of the corresponding author (i.e., the author to whom notices will be sent concerning the competition): Gustavo Olague ------------------------------------------------------------------------------------------------------ 4. The abstract of the paper(s): Nowadays, object recognition is widely studied under the paradigm of matching local features. This work describes a genetic programming methodology that synthesizes mathematical expressions that are used to improve a well known local descriptor algorithm. It follows the idea that object recognition in the cerebral cortex of primates makes use of features of intermediate complexity that are largely invariant to change in scale, location, and illumination. These local features have been previously designed by human experts using traditional representations that have a clear, preferably mathematically, well-founded definition. However, it is not clear that these same representations are implemented by the natural system with the same representation. Hence, the possibility to design novel operators through genetic programming represents an open research avenue where the combinatorial search of evolutionary algorithms can largely exceed the ability of human experts. This paper pro vides evidence that genetic programming is able to design new features that enhance the overall performance of the best available local descriptor. Experimental results confirm the validity of the proposed approach using a widely accept testbed and an object recognition application. ------------------------------------------------------------------------------------------------------ 5. A list containing one or more of the eight letters (A, B, C, D, E, F, G, or H) that correspond to the criteria (see above) that the author claims that the work satisfies: (A) The result was patented as an invention in the past, is an improvement over a patented invention, or would qualify today as a patentable new invention. (B) The result is equal to or better than a result that was accepted as a new scientific result at the time when it was published in a peer-reviewed scientific journal. (C) The result is equal to or better than a result that was placed into a database or archive of results maintained by an internationally recognized panel of scientific experts. (D) The result is publishable in its own right as a new scientific result 3/4 independent of the fact that the result was mechanically created. (E) The result is equal to or better than the most recent human-created solution to a long-standing problem for which there has been a succession of increasingly better human-created solutions. (F) The result is equal to or better than a result that was considered an achievement in its field at the time it was first discovered. (G) The result solves a problem of indisputable difficulty in its field. ------------------------------------------------------------------------------------------------------ 6. A statement stating why the result satisfies the criteria that the contestant claims: The following states why our entry satisfies criteria A,B,C,D,E,F and G: (A) The result was patented as an invention in the past, is an improvement over a patented invention, or would qualify today as a patentable new invention. Our proposed methodology for synthesizing descriptor operators represent an improvement over a patented descriptor algorithm called SIFT (Scale Invariant Feature Transform). The idea was to find through GP a set of mathematical expressions that could be equal or better than the weighted gradient magnitude that is applied within the SIFT descriptor. We proposed to name these mathematical expressions as descriptor operators. The SIFT patent is the following: "Method and apparatus for identifying scale invariant features in an image and use of same for locating an object in an image". David G. Lowe, US Patent 6,711,293 (March 23, 2004). Asignee: The University of British Columbia. (B) The result is equal to or better than a result that was accepted as a new scientific result at the time when it was published in a peer-reviewed scientific journal. Numerous local descriptors have been published in scientific journals since the idea first appeared as a new paradigm for several computer vision applications. Here, we compare our results with previous published descriptors for which their evaluation technique is based on a recall vs 1-precision space. Thus, we test several works to compare our descriptor algorithm and in particular we find that our results surpassed the overall performance of previous local descriptors including the following: [1] David G. Lowe, "Distinctive image features from scale-invariant keypoints," International Journal of Computer Vision, 60(2):91-110, 2004. [2] K. Mikolajczyk, C. Schmid, A performance evaluation of local descriptors. IEEE Transactions on Pattern Analysis and Machine Learning, 27(10):1615-1630, 2005. [3] Herbert Bay, Andreas Ess, Tinne Tuytelaars, Luc Van Gool, "SURF: Speeded Up Robust Features", Computer Vision and Image Understanding (CVIU), 110(3):346-359, 2008. (C) The result is equal to or better than a result that was placed into a database or archive of results maintained by an internationally recognized panel of scientific experts. We used a testbed that is widely accepted as a standard performance evaluation for local descriptors in the computer vision community. This testbed was originally developed by researchers from INRIA Rhone Alpes and the current test includes contributions from University of Oxford, Katholieke Univesiteit Leuven, and the Center for Machine Perception at the Czech Technical University. Today, it is mantained by the Visual Geometry Group of the Robotics Research Group and it is available at the following address: http://www.robots.ox.ac.uk/~vgg/research/affine/ As a matter of fact, in order to use this testbest in our GP framework, we propose to include the F-Measure in the evaluation process to obtain not only a graphical result as is commonly performed, but also a cuantitative measure as required in a GP optimization framework. (D), (E) and (F) (D) The result is publishable in its own right as a new scientific result 3/4 independent of the fact that the result was mechanically created. (E) The result is equal to or better than the most recent human-created solution to a long-standing problem for which there has been a succession of increasingly better human-created solutions. (F) The result is equal to or better than a result that was considered an achievement in its field at the time it was first discovered. Our methodology for automatically obtaining new descriptor operators using genetic programming represents a new approach within the computer vision field; in particular, it address a new approach where local descriptors could be synthesized through GP. As a by product, the results found by genetic programming in the experimental stage surpassed our initial expectations; indeed, we obtained much better performance than the human-made descriptor algorithms. As a conclusion, we have improved the SIFT algorithm which has been considered until now, an achievement in its field using GP. (G) The result solves a problem of indisputable difficulty in its field. Today, most computer vision conferences and journals devote a special session or section to local descriptors research because it has became a powerful technique for solving real-world vision problems. Thus, our proposed technique opens a research avenue towards evolutionary learning of local descriptors. Here, we demostrated the effectiveness of our GP approach through an extensive experimental study and its application using an object recognition problem. ------------------------------------------------------------------------------------------------------ 7. A full citation of the paper (that is, author names; publication date; name of journal, conference, technical report, thesis, book, or book chapter; name of editors, if applicable, of the journal or edited book; publisher name; publisher city; page numbers, if applicable); Perez C.B., Olague G. "Learning Invariant Region Descriptor Operators with Genetic Programming and the F-measure". International Conference on Pattern Recognition. Pages 1-4. ISBN: 978-1-4244-2174-9. December 8-11, 2008. Perez C.B., Olague G. "Evolutionary Learning of Local Descriptor Operators for Object Recognition". To appear in Genetic and Evolutionary Computation Conference. July 8-12, 2009. ------------------------------------------------------------------------------------------------------ 8. A statement either that "any prize money, if any, is to be divided equally among the co-authors" OR a specific percentage breakdown as to how the prize money, if any, is to be divided among the co-authors: Any prize money is to be divided equally among the co-authors. ------------------------------------------------------------------------------------------------------ 9. A statement stating why the judges should consider the entry as "best" in comparison to other entries that may also be "human-competitive." In this work, besides opening a new research avenue towards the synthesis of local descriptors, we believe that this kind of formulation shows a rigorous path in the design of computer vision applications where genetic programming plays a major role; thus strengthening the emerging area of evolutionary computer vision[1,2]. In particular, our proposal defines the problem in terms of machine learning search in which no-prior method exists for generating mathematical expressions that correspond to descriptor operators. In order to tackle such idea, we propose the F-measure as a cuantitative measure to evaluate local descriptors. Even though such metric was already known the approach of GP allows us to propose it in order to frame correctly the optimization process. The experiments confirm that GP improves the SIFT descriptor with impacting results. [1] Cagnoni S., Lutton E., and Olague G. (eds). Evolutionary Computer Vision. Evolutionary Computation. MIT Press. Vol. 26, No.4, Pages 437-438. 2008. [2] Olague G., Lutton E., and Cagnoni S. (eds). Introduction to the Special Issue on Evolutionary Computer Vision and Image Understanding. Pattern Recognition Letters, Vol.27, No.11, Pages 1161-1163. 2006.