A carregar...
Publicação
An API For Building Artificial Worlds For Machine Learning Using Blender
| Nome: | Descrição: | Tamanho: | Formato: | |
|---|---|---|---|---|
| 2.06 MB | Adobe PDF |
Orientador(es)
Resumo(s)
Virtual technologies are a facet of advancement that permeates almost all major fields in our world,
from physics simulations to videogames and movies, urban design and engineering to rocket science,
all make use of this wonderful and powerful tool. One such field is Machine Learning. This project aims
to leverage said virtual technologies to help such a field in the shape of an object identification model
being developed by Thibault Nicolas Langlois.
This is done by using a tool such as Blender and a language such as Python to create an API that
leverages both in order to facilitate the procedural generation of virtual worlds. By generating said
worlds and taking snapshots of them with the rendering power of Blender, this project aims to help
generate vast amounts of data that can be catalogued and sent to feed said models.
To do this, the API was constructed in a way that allows for modular adaptation to whatever purpose it
is to be used for, all while containing an example focused on city generation. This can help guide others
intending on using the API, on how to adapt it to their use case.
This proved to be extremely complex and arduous, requiring many hours of work and lines of code
written. The scale of the endeavour was vast and required a lot to reach the state it currently presents
itself in. It also presented how capable and versatile the technology can be, giving many a tool to the
user’s disposal, as well as the developer.
This proves the path forward for testing many aspects of products or experiments is indeed virtualization
as it has shown to be capable enough in providing the tools necessary to emulate even the most complex
scenarios, all while potentially saving numerous amounts of investment and time, depending on the scale
of the products taking advantage of it.
Descrição
Trabalho de projeto de mestrado, Informática, 2022, Universidade de Lisboa, Faculdade de Ciências
Palavras-chave
Blender Python Procedural Generation Scripting Virtual Worlds Trabalhos de projeto de mestrado - 2023