Feature learning for natural images annotation

• Grant number: 13/04172-0
• Support type: Scholarships in Brazil - Post-Doctorate
• Effective date (Start): July 01, 2013
• Effective date (End): May 31, 2014
• Field of knowledge: Physical Sciences and Mathematics - Computer Science - Computing Methodologies and Techniques
• Principal Investigator: Alexandre Xavier Falcão
• Grantee: David Menotti Gomes
• Home Institution: Instituto de Computação (IC). Universidade Estadual de Campinas (UNICAMP). Campinas , SP, Brazil

Abstract

Natural image annotation (photos, videos) by assigning keywords that describe the scene content makes the automatic image indexing feasible and allows further efficient content based image retrieval. Large image databases makes manual image annotation an unfeasible task. The traditional approach consists in manually annotate some images and then, by using these images, train a supervised pattern classifier to automatically annotate the remaining ones. The advisor's research group has investigated, through a PhD thesis, active learning techniques for the selection of the most representative images for manual annotation. Nonetheless, the classification effectiveness mainly depends on the image descriptor/feature used in this process.This project, therefore, has as main goals the study and design of deep learning techniques for image description. These techniques are usually based on sequences of operations such as normalization, filtering, and selection (pooling) at some scales. At each scale, linear filter banks are applied on image and their coefficients can be generated in a randomly way or in a more elaborated one. In the latter case, unsupervised learning algorithms are usually applied in region samples obtained from the image database.The project will investigate those and other learning techniques of these filter coefficients, also taking into account the annotation of some training images, selection techniques, and association rules of descriptors to regions of the images, obtained from segmentation techniques of the images in superpixels.

Scientific publications

(References retrieved automatically from Web of Science and SciELO through information on FAPESP grants and their corresponding numbers as mentioned in the publications by the authors)

MENOTTI, DAVID; CHIACHIA, GIOVANI; PINTO, ALLAN; SCHWARTZ, WILLIAM ROBSON; PEDRINI, HELIO; FALCAO, ALEXANDRE XAVIER; ROCHA, ANDERSON. Deep Representations for Iris, Face, and Fingerprint Spoofing Detection. IEEE Transactions on Information Forensics and Security, v. 10, n. 4, p. 864-879, APR 2015. Web of Science Citations: 136.