Abstract
The ever-growing capabilities of computers have enabled pursuing Computer Vision through Machine Learning (i.e., MLCV). ML tools require large amounts of information to learn from (ML datasets). These are costly to produce but have received reduced attention regarding standardization. This prevents the cooperative production and exploitation of these resources, impedes countless synergies, and hinders ML research. No global view exists of the MLCV dataset tissue. Acquiring it is fundamental to enable standardization. We provide an extensive survey of the evolution and current state of MLCV datasets (1994 to 2019) for a set of specific CV areas as well as a quantitative and qualitative analysis of the results. Data were gathered from online scientific databases (e.g., Google Scholar, CiteSeerX). We reveal the heterogeneous plethora that comprises the MLCV dataset tissue; their continuous growth in volume and complexity; the specificities of the evolution of their media and metadata components regarding a range of aspects; and that MLCV progress requires the construction of a global standardized (structuring, manipulating, and sharing) MLCV “library”. Accordingly, we formulate a novel interpretation of this dataset collective as a global tissue of synthetic cognitive visual memories and define the immediately necessary steps to advance its standardization and integration.

Abstract
The FotoInMotion (FiM) project is building a novel media creation platform, leveraging the use of semi-automated analysis and editing tools to empower creators to easily transform static visual acquisitions of real-world events into rich, animated and engaging objects, distributable through common channels. FiM transforms the content creative chain into an integrated pipeline across which media and metadata seamlessly flow and are exploited to produce more complex media objects. One of the addressed challenges consists the need for a seamless and efficient communication across such pipeline and on how to preserve, in a structured manner, all of the involved media and metadata. Existing standardized metadata tools and content wrappers are limited in expressivity and scope and incapable of fully supporting the needs of the content creative pipeline. This paper describes FiM's new structured data object, i.e. the Digital Event (DE), which acts as a universal vehicle for media and metadata. It builds on well-established and emergent MPEG standards (MPEG-21, MPEG-V, MPEG-7 and MPEG HEIF), to support data diversity, interoperability, packaging and sharing, within complex, Machine Learning enhanced, creative pipelines. Our solution has been validated by creative professionals (photojournalism, fashion marketing and festivals), who have conducted experiments within the context of different creative workflows in real world scenarios. DE's employment revealed to be advantageous, particularly in the homogenization of the media and metadata representation and packaging and in the normalization of the interaction between different pipeline components.

Abstract
Machine Learning (ML), presently the major research area within Artificial Intelligence, aims at developing tools that can learn, approximately on their own, from data. ML tools learn, through a training phase, to perform some association between some input data and some output evaluation of it. When the input data is audio or visual media (i.e. akin to sensory information) and the output corresponds to some interpretation of it, the process may be described as Synthetic Cognition (SC). Presently ML (or SC) research is heterogeneous, comprising a broad set of disconnected initiatives which develop no systematic efforts for cooperation or integration of their achievements, and no standards exist to facilitate that. The training datasets (base sensory data and targeted interpretation), which are very labour intensive to produce, are also built employing ad-hoc structures and (metadata) formats, have very narrow expressive objectives and thus enable no true interoperability or standardisation. Our work contributes to overcome this fragility by putting forward: a specification for a standard ML dataset repository, describing how it internally stores the different components of datasets, and how it interfaces with external services; and a tool for the comprehensive structuring of ML datasets, defining them as Synthetic Cognitive Experience (SCE) records, which interweave the base audio-visual sensory data with multilevel interpretative information. A standardised structure to express the different components of the datasets and their interrelations will promote re-usability, resulting on the availability of a very large pool of datasets for a myriad of application domains. Our work thus contributes to: the universal interpretability and reusability of ML datasets; greatly easing the acquisition and sharing of training and testing datasets within the ML research community; facilitating the comparison of results from different ML tools; accelerating the overall research process. © MIR Labs.

Abstract
Over the last decade noticeable progress has occurred in automated computer interpretation of visual information. Computers running artificial intelligence algorithms are growingly capable of extracting perceptual and semantic information from images, and registering it as metadata. There is also a growing body of manually produced image annotation data. All of this data is of great importance for scientific purposes as well as for commercial applications. Optimizing the usefulness of this, manually or automatically produced, information implies its precise and adequate expression at its different logical levels, making it easily accessible, manipulable and shareable. It also implies the development of associated manipulating tools. However, the expression and manipulation of computer vision results has received less attention than the actual extraction of such results. Hence, it has experienced a smaller advance. Existing metadata tools are poorly structured, in logical terms, as they intermix the declaration of visual detections with that of the observed entities, events and comprising context. This poor structuring renders such tools rigid, limited and cumbersome to use. Moreover, they are unprepared to deal with more advanced situations, such as the coherent expression of the information extracted from, or annotated onto, multi-view video resources. The work here presented comprises the specification of an advanced XML based syntax for the expression and processing of Computer Vision relevant metadata. This proposal takes inspiration from the natural cognition process for the adequate expression of the information, with a particular focus on scenarios of varying numbers of sensory devices, notably, multi-view video.

Abstract
The Web is rapidly becoming the prime medium for human socialization. The resources that enable that process (social web sites, blogs, media objects, etc.) present growing complexity and, collectively, weave an ever more intricate web of relationships. Current technology for declaring those relationships is predominantly implicit, ambiguous and semantically poor. As a consequence, their automatic assessment is complex and error prone, preventing the satisfaction of users' needs such as effective semantic searches. To address these limitations, whilst enabling the explicit declaration of semantically unambiguous relationships between digital resources, a solution employing structured semantic descriptors and ontologies was conceived, based on MPEG-21. This paper explains the functioning of the devised mechanism, and goes beyond that, into the definition of two novel employment venues for it, at the service of two real-world usage scenarios. These demonstrate the mechanism's added value as a powerful alternative for the semantically aware interconnection of web resources, and highlight the increased QoE that said mechanism enables.