A Novel Graph Embedding Framework for Object Recognition

Abstract

In recent years research has been producing an important effort to encode the digital image content. Most of the adopted paradigms only focus on local features and lack in information about location and relationships between them. To fill this gap, we propose a framework built on three cornerstones. First, ARSRG (Attributed Relational SIFT (Scale-Invariant Feature Transform) regions graph), for image representation, is adopted. Second, a graph embedding model, with purpose to work in a simplified vector space, is applied. Finally, Fast Graph Convolutional Networks perform classification phase on a graph based dataset representation. The framework is evaluated on state of art object recognition datasets through a wide experimental phase and is compared with well-known competitors.

Bag of ARSRG Words (BoAW)

“…In Table 1 we present only the results considering a batch of 400 images, since intermediate results do not provide particular improvements. We show the results achieved by BoVW and those obtained in [24] using some variants of linear discriminant analysis (ILDAaPCA, batchLDA, ILDAonK and ILDAonL) and in [25] (ARSRGemb). ILDAaPCA creates a PCA subspace using the k dimensional reconstructive subspace and c − 1 additional vectors having discriminative features.…”

“…In order to obtain a valid comparison with the methods in [26,27] we adopted the same settings: 25 objects are randomly selected and 11% are used as the training set and 89% are used as the testing set. Therefore, results obtained by BoVW are shown and those obtained in [26,27] by applying their solution (VFSR) and the approaches proposed in [28] (gdFil), in [29] (APGM), in [30] (VEAM), in [31] (DTROD-AdaBoost), in [32] (RSW+Boosting), in [33] (Sequential Patterns), in [34] (LAF) and in [25] (ARSRGemb). Results are listed in form of average accuracy and the approach that provided the best performance is highlighted.…”

“…The testing set is composed of 60 per category (15 views per object). We present tests performed by BoVW and those achieved in [26] by employing the solution proposed in [25] (ARSRGemb), [28] (gdFil), in [29] (APGM), and in [30] (VEAM). Also in this case the results are listed highlighting the accuracy of the best approach.…”

DOPSIE: Deep-Order Proximity and Structural Information Embedding

“…In the last decade, we have observed an exponential increase of real network datasets. Networks (or graphs) have been adopted to encode information in different fields such as: computational biology [1], social network sciences [2], computer vision [3,4], and natural-language processing [5].…”

Attributed Relational SIFT-Based Regions Graph for Art Painting Retrieval

“…Other systems, called Region Based Image Retrieval (Liu, Zhang, Lu, & Ma, 2007) (RBIR), focus their attention on specific image regions instead of the entire content to extract features. In this paper, a graph structure for image representation, called Attributed Relational SIFT-based Regions Graph (ARSRG), is described, analyzed and discussed with reference to previous works (Manzo & Petrosino, 2013;Manzo et al, 2014;Manzo & Pellino, 2019;Manzo, 2019). In particular, new definitions and properties arising from the detailed analysis of the structure are introduced.…”