Towards an Images Dataset Processing trough Supervised and Unsupervised Learning

Abstract

Internet offers to its users an ever-increasing number of information. Among those, the multimodal data (images, text, video, sound) are widely requested by users, and there is a strong need for effective ways to process and to manage it, respectively. Most of existed algorithms/frameworks are doing only images annotations and the search is doing by these annotations, or combined with some clustering results, but most of them do not allow a quick browsing of these images. Even if the search is very quickly, but if the number of images is very large, the system must give the possibility to the user to browse this data. In this paper we investigate the use of the supervised learning to classify an images dataset and the unsupervised learning to browse the images. In our proposed schema, we used both PCA and LDA to transform the feature space and then to classify the dataset. We used this technique for all five datasets available on the challenge web site of The German Traffic Sign Recognition Benchmark: HOG1, HOG2, HOG3, HueHIst and Haar [7]. Finnaly we used a voting approach to find the consensus for all five partitions. Also, an application to the images browsing is shown using the topological unsupervised learning.