Image and Video Understanding in Big Data

h 1 The huge volume of data produced every day is posing a signifiant challenge to computer scientists since it is infeasible that this ata can be effectively processed and consistently interpreted by umans manually, even to a very small extent. Due to the automaion of many industrial processes and the advent of cheaper and igh-performance sensors, many aspects of life, including medial, commercial, industrial and security areas are increasingly more haracterized by large data collections that must also be processed nd communicated with. DNA sequencing, radiographic and imagng examinations (e.g. MRI, X Ray, echography), urban surveillance, raffic monitoring, customer profiling and e-commerce, visual inpection, industrial machine maintenance and failure prediction re some examples of the vast range of possible applications in ur everyday social and working life characterized by large data atherings. Among these, visual data takes a prominent role given the large nd pervasive diffusion of imaging devices in our cities, industries, t home, and in our hands given the plenitude of personal deices such as smart phones at our disposal. For the latter, one can ppreciate readily the scale of the data consumed just by thinkng about the amount of image and video data downloaded evry minute in social media such as Facebook, Instagram, Snapchat, mong others. Moreover, visual data is by large the most diverse nd demanding media, as compared to text for instance, growing t an unprecedented speed. This requires the design of effective ethods to manage it, by mining relevant information while disarding redundant or useless data. To that end, more scalable and obust methods are required to efficiently index, retrieve, organize, nterpret and interact with such big visual data, and this cannot be one without automatic or semi-automatic, e.g. human in the loop, ethods and processes capable of distil useful observations from a arge quantity of raw data. In this context, it is clear that big visual data analysis and unerstanding impose significant scientific and technological chalenges since one not only should have to advance methods able to racefully scale to both big and diverse data whilst being computaionally cheap, but also should need to develop processes suitable or learning from big and diverse data without incurring prohibitive osts in human and monetary resources, and of time, required y exhaustive data annotation. Moreover, efficient online learning ethods could be required to cope with data acquired over time.