Examensarbete Adam Larsson
Contact: magnus [dot] oskarsson [at] math [dot] lth [dot] se
Save event to your calendar
Adam Larsson (F17) presenterar sitt examensarbete Methods for generating façade sketches from photographs on mobile devices (Svensk titel: Metoder för generering av fasadritningar från fotografier på mobila enheter)
Technology has come far in our modern society, but many smaller companies in certain lines of work still do not make use of the powerful methods discovered in image analysis and neural networks in recent years. The purpose of this report is to implement an algorithm which can generate sketches of house façades from images in order to streamline the façade inspection process. The paper experiments with classical image analysis methods as well as neural network solutions, such as DexiNed for edge detection and MobileNet for segmentation, with the end goal of finding an efficient algorithm which can be run on a mobile device. The findings point towards using a combination of a filter-algorithm and an augmented version of the DexiNed network to produce realistic sketches with a pencil drawing-like aesthetic which can be run on a mobile device producing good looking sketches with a very low inference time.
Oscar Bark (BIT ADDICT), Magnus Oskarsson (Matematikcentrum)
Niels Christian Overgaard (Matematikcentrum)