Big Visual Data Analysis: Scene Classification and Geometric Labeling (SpringerBriefs in Electrical and Computer Engineering)
Format: PDF / Kindle (mobi) / ePub
This book offers an overview of traditional big visual data analysis approaches and provides state-of-the-art solutions for several scene comprehension problems, indoor/outdoor classification, outdoor scene classification, and outdoor scene layout estimation. It is illustrated with numerous natural and synthetic color images, and extensive statistical analysis is provided to help readers visualize big visual data distribution and the associated problems. Although there has been some research on big visual data analysis, little work has been published on big image data distribution analysis using the modern statistical approach described in this book. By presenting a complete methodology on big visual data analysis with three illustrative scene comprehension problems, it provides a generic framework that can be applied to other big visual data analysis tasks.
Grouping Using Experts’ Decisions Data grouping is a widely adopted technique in statistics and machine learning. As millions of visual data, such as images and video clips, are created with the emergence of smart phones and mobile devices every day, an efficient data grouping scheme is important to visual data analytics. Besides, understanding the capabilities of machine-trained experts from the big visual data distribution aspect is crucial to an integrated smart system. In this section, we
1961–1968. IEEE (2011) 14. Han, F., Zhu, S.C.: Bottom-up/top-down image parsing with attribute grammar. IEEE Trans. Pattern Anal. Mach. Intell. 31(1), 59–73 (2009) 15. Hedau, V., Hoiem, D., Forsyth, D.: Recovering the spatial layout of cluttered rooms. In: 2009 IEEE 12th International Conference on Computer vision, pp. 1849–1856. IEEE (2009) 16. Hoiem, D., Efros, A., Hebert, M.: Closing the loop in scene interpretation. In: IEEE Conference on Computer Vision and Pattern Recognition, 2008. CVPR
contain more than one object, and users are allowed to label these objects. Finally, its numbers of images and object classes can be easily increased. 14 2 Scene Understanding Datasets Fig. 2.7 Examples in the ImageNet dataset 2.2.5 Scene Understanding (SUN) Dataset The Scene understanding (SUN) dataset, introduced by Xiao et al., finds applications in many research fields, such as scene recognition, computer vision, human perception, cognition and neuroscience, machine learning, data
Images , Line Features [22, 23], Texton Histograms , Color Histograms, Geometric Probability Map  and Geometric specific histograms . The extracted features are all relevant to scene classification. The classifier is trained using the one-vs-remaining SVM. To compare the performance on different datasets, experiments are also conducted on the 15-scene dataset [1, 3, 4]. The results are shown in Fig. 2.10. The performance curve labeled by “all” is to adopt the weighted sum of
3485–3492. IEEE (2010) 15. Gao, T., Koller, D.: Discriminative learning of relaxed hierarchy for large-scale visual recognition. In: 2011 IEEE International Conference on Computer Vision (ICCV), pp. 2072–2079. IEEE (2011) 16. Pavlopoulou, C., Yu, S.X.: Indoor-outdoor classification with human accuracies: image or edge gist? In: 2010 IEEE Computer Society Conference on Computer Vision and Pattern Recognition Workshops (CVPRW), pp. 41–47. IEEE (2010) 17. Griffin, G., Holub, A., Perona, P.: