Akka Zemmari - Deep Learning in Mining of Visual Content

SpringerBriefs present concise summaries of cutting-edge research and practical applications across a wide spectrum of fields. Featuring compact volumes of 50 to 125 pages, the series covers a range of content from professional to academic.

Briefs allow authors to present their ideas and readers to absorb them with minimal time investment. Briefs will be published as part of Springers eBook collection, with millions of users worldwide. In addition, Briefs will be available for individual print and electronic purchase. Briefs are characterized by fast, global electronic dissemination, standard publishing contracts, easy-to-use manuscript preparation and formatting guidelines, and expedited production schedules. We aim for publication 812 weeks after acceptance. Both solicited and unsolicited manuscripts are considered for publication in this series.

More information about this series at http://www.springer.com/series/10028

This Springer imprint is published by the registered company Springer Nature Switzerland AG.

The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland

Deep learning is a new trend in artificial intelligence pushing the performances of machine learning algorithms to their real-life applicability for solving a variety of problems the humanity faces nowadays.

Computer vision and multimedia indexing research have realized a general move from all previous approaches to those on the basis of deep learning.

This approach is built on the principles of supervised machine learning and in particular, artificial neural networks. Applied to visual information mining, it also incorporates our fundamental knowledge in image processing and analysis. Today, we translate all our know-how in visual information mining into this language.

The proliferation of software frameworks allows for easy design and implementation of deep architectures, for the choice and adequate parameterization of different optimization algorithms for training parameters of deep neural networks. The availability of graphical processing units (GPU) and of distributed computing made the computational times for learning quite reasonable. For young researchers and those who move to this kind of methods it is important, we think, to get very quickly into comprehension of underlying mathematical models and formalism, but also to make a bridge between the methods previously used for understanding of images and videos and these winning tools.

It is difficult today to write a book about deep learning, so numerous are different tutorials easily accessible on the Internet. What is the particularity of our book compared to them? We tried to keep a sufficient balance between the usage of mathematical formalism, graphical illustrations, and real-world examples. The book should be easy to understand for young researchers and professionals with engineering and computer science background. Deep learning without pain, this is our goal.

We thank our PhD students, Abraham Montoya Obeso and Karim Aderghal, for their kind help in preparation of illustrative material for this book.