Detekcija pomorskih markera korištenjem kamera

undergraduate thesis

• Title (english): Detection of sea markers using cameras
• Author: Maja Magdalenić
• Mentor: Nikola Mišković (mentor)
• Committee member: Nikola Mišković (predsjednik povjerenstva)
• Committee member: Ivan Lončar (član povjerenstva)
• Committee member: Nadir Kapetanović (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; Zagreb
• Defense date and country: 2021-09-14, Croatia
• Scientific / art field, discipline and subdiscipline: TECHNICAL SCIENCES; Computing

Abstract

U ovom radu najprije je opisan način rada detektora objekata, a zatim detaljno objašnjen postupak izrade detektora objekata (pomorskih markera) korištenjem Tensorflow platforme za strojno učenje i Tensorflow Object Detection alata za korištenje unaprijed treniranih modela metodom prijenosa učenja. Razvijena su 2 modela detektora različite arhitekture (Faster R-CNN i SSD). Modeli su validirani pomoću COCO metrike te je dan grafički prikaz rezultata. Također, oba modela su testirana na određenom setu podataka, a na temelju toga napravljena je usporedba performansi modela te je donesen zaključak koji bi model bio prikladniji za korištenje.

Abstract (english)

This paper first describes how object detectors work and then explains in detail the process of making object detector (in this case sea markers) using Tensorflow deep learning platform and Tensorflow Object Detection tool for using pre-trained models with learning transfer method. Two models of detectors with different architecture (Faster R-CNN and SSD) were developed. The models were validated using COCO metrics and a graphical representation of the results was given. Also, both models were tested with a specific data set, and based on that, a comparison of the model’s performances was made. In the end there is a conclusion which model would be more appropriate.

• Keywords: duboko učenje; detekcija objekata; neuronske mreže; pomorski markeri; klasifikacija; lokalizacija; set podataka; unaprijed obučeni model
• Keywords (english): deep learning; object detection; neural networks; sea markers; classification; localization; data set; pre-trained model
• Language: croatian
• URN:NBN: urn:nbn:hr:168:689310
• Study programme: Title: Computing; Study programme type: university; Study level: undergraduate; Academic / professional title: sveučilišni/a prvostupnik/ prvostupnica (baccalaureus/baccalaurea) inženjer/inženjerka računarstva (sveučilišni/a prvostupnik/ prvostupnica (baccalaureus/baccalaurea) inženjer/inženjerka računarstva)
• Type of resource: Text
• File origin: Born digital
• Access conditions: Closed access
• Created on: 2023-01-14 09:20:17