Copia de Copia CUCU de La regin sin nosotros / tesis MAEDI con modificaciones 12-1
This research studies the development and possibilities of video editing through the use of Artificial Intelligence, in the context of systems for moderating and distributing video content, mainly on the YouTube platform. The goal of this research is to recognize and project some of the expressive c...
Guardado en:
| Autor principal: | |
|---|---|
| Otros Autores: | |
| Formato: | Tesis de maestría acceptedVersion |
| Lenguaje: | Español |
| Publicado: |
Universidad de Buenos Aires. Facultad de Arquitectura, Diseño y Urbanismo
2024
|
| Materias: | |
| Acceso en línea: | http://repositoriouba.sisbi.uba.ar/gsdl/cgi-bin/library.cgi?a=d&c=aaqmas&cl=CL1&d=HWA_7681 https://repositoriouba.sisbi.uba.ar/gsdl/collect/aaqmas/index/assoc/HWA_7681.dir/7681.PDF |
| Aporte de: |
| Sumario: | This research studies the development and possibilities of video editing through the use of Artificial Intelligence, in the context of systems for moderating and distributing video content, mainly on the YouTube platform. The goal of this research is to recognize and project some of the expressive capabilities that have arisen from the computerization of the vast amount of video content circulating on the web, and at the same time to reveal a portion of the logic that operates the technology that gives order to the available audiovisual material, associating this logic of order with that of cinematographic editing/montage. This research is framed in the technical understanding of the algorithms and artificial neural networks that drive the computer description of videos, so the focus is on technical bibliography during the thesis.
On the other hand, works of found footage cinema and structural cinema are also contrasted with some of the possibilities of automated video editing, and also texts by authors such as Sergei Eisenstein are used, whose studies on editing come into tension with the aforementioned genres of experimental cinema.
Within the methodological resources, historical intersections are generated regarding the order and juxtaposition of images, associated with media genealogies, including devices from the 19th century in the study, comparing them with other current developments such as recent text to image (text2image) technique and automated video editing applications (neural edit)
Among the findings of the research is the fact that the labeling and moderation tools developed to organize and distribute the huge amount of videos circulating on the network (a quantity whose volume corresponds to the magnitudes called Big Data), are adapted as part of the machinery used to understand and assemble sequences in the automated video assembly applications using AI used by millions of users.
It is reasonable to conclude that the generalist cut introduced by these models trained on the web, used primarily to suggest purchases and censor inappropriate content, can make little contribution if they are put to work as editors, in the classic sense and utility of the term, that is, the one that implies emulating human sensitivity, creating sense by juxtaposing images and sounds.
So, another form of editing is then proposed for these same tools, developing other types of algorithms, driven by these same models and neural networks, but carrying out free searches in the audiovisual universe, in a kind of audiovisual mining. These algorithms could edit video sequences, anthropologically valuable, opening up to general knowledge the human routines and attitudes recorded on video, crossing all cultures and ethnicities, being an unprecedented mirror of what people produce in different parts of the planet, and at the same time putting AI, and computing in a complementary relationship with that of human beings, without replacing people in their sensitive domain, but helping them to better understand the world. |
|---|