Prepoznava plovil na reki

BESEDNJAK, ŽAN (author), Šajn, Luka (mentor)

Cilj diplomske naloge je ustvariti sistem za analizo rečnega prometa in turizma s štetjem specifičnih plovil. Model bo osnovan na družini modelov za prepoznavo YOLO. Na izbiro je vplivala predvsem zmožnost YOLO modelov za prepoznavo v realnem času - več kot 60 slik na sekundo. Podatki za učenje in ocenjevanje bodo pridobljeni s snemanjem prometa na reki Soči na različnih lokacijah in v različnih vremenskih razmerah. Iz potrebe po veliki količini označenih podatkov bomo uporabili metode za avtomatsko označevanje. Ocenjevanje bo izvedeno na testnih podatkih in v realnem primeru uporabe.

• Language: Slovenian
• Keywords: Prepoznava, Označevanje, YOLO, rečni promet
• Work type: Bachelor thesis/paper
• Organization: FRI - Faculty of Computer and Information Science
• Year: 2024
• PID: 20.500.12556/RUL-155941
• Publication date in RUL: 24.04.2024

Secondary language

• Language: English
• Title: River Vessels Recognition

Abstract

The purpose of this thesis is to create a model for recognition of river vessels. It is going to be used for counting different vessels and analyzing river traffic and tourism activities. The model will be based on a family of YOLO models for recognition. The most important factor in the decision was the property of YOLO models to recognize objects in real time - more than 60 frames per seconds. Data for learning and evaluation is going to be acquired with filming traffic on river Soča on different locations and weather conditions. In need for a large set of labeled data we use systems for automatic labeling. Evaluation will be done on the test data and on real world examples.

• Keywords: Recognition, Labeling, YOLO, river traffic