3D modelling with a webcam
Typ
Examensarbete på kandidatnivå
Program
Publicerad
2021
Författare
Nevalainen Henaes, Axel
Rehnberg, Robin
Modellbyggare
Tidskriftstitel
ISSN
Volymtitel
Utgivare
Sammanfattning
Creating convincingly realistic three dimensional models is a common endeavour in fields like art and entertainment. The process can be both time consuming and difficult. By reconstructing objects from photographs this process could be sped up significantly while also lowering the barrier of entry when it comes to producing realistic models. The aim of this project is to explore such a solution by attempting
to introduce the third dimension, depth, into images.
This project explores and research the different methodologies that can used to recre ate three dimensional environments using different mathematical concepts, together with algorithms and possibly artificial intelligence
The end result was a success where the program successfully managed to introduce the third dimension into images. This depth is then used to create three dimensional model. Because this was successfully done with enough time left, machine learning was introduced to the project to compare the results from the algorithmic approach to that of machine learning. The results from the machine learning bring a level
of detail to the three dimensional images that can not be matched by algorithmic solutions.
Further development of the project would include merging information provided by the photos from multiple angles. In its current state it works for objects with a known viewing angle for the context it will be used in. But with data captured from different angles merged together such preparations would not have to be made.