diplomsko delo
Jernej Avsec (Author), Matej Kristan (Mentor), Vitjan Zavrtanik (Co-mentor)

Abstract

V zadnjem času je bil narejen izjemen napredek na področju štetja objektov z malo učnimi primeri. Trenutni najuspešnejši globalni števci temeljijo na napovedovanju gostotnih map, katerih vsota je ocena števila objektov. Velika pomanjkljivost teh metod pa je, da rezultati niso razložljivi, saj ne podajo tudi lokacije objektov, kar je ključno za številne aplikacije. V diplomski nalogi predlagamo metodo CVDFC, ki s pomočjo difuzijskega modela izboljšuje kakovost gostotnih map in jih pretvori v natančne lokacijske točke objektov. Predlagani model uporablja pogojni difuzijski proces za generiranje lokacijskih točk, modul za štetje objektov pa izvede tlačenje lokalnih ne-maksimumov (NMS) nad generiranimi točkami, kar omogoča natančno štetje in lokalizacijo objektov na sliki. Eksperimentalni rezultati so pokazali, da metoda CVDFC pri nalogi štetja objektov za 30% prekaša referenčno metodo LOCA kombinirano z lokalizacijo objektov preko NMS. Prav tako se je CVDFC izkazala za konkurenčno glede na druge metode, kar kaže na njeno učinkovitost in praktično uporabnost na področju štetja objektov z malo primeri.

Keywords

difuzijski modeli;štetje objektov;nevronske mreže;lokacijske točke;gostotna mapa;računalništvo;univerzitetni študij;diplomske naloge;

Data

Language: Slovenian
Year of publishing:
Typology: 2.11 - Undergraduate Thesis
Organization: UL FRI - Faculty of Computer and Information Science
Publisher: [J. Avsec]
UDC: 004.93(043.2)
COBISS: 211893507 Link will open in a new window
Views: 135
Downloads: 44
Average score: 0 (0 votes)
Metadata: JSON JSON-RDF JSON-LD TURTLE N-TRIPLES XML RDFA MICRODATA DC-XML DC-RDF RDF

Other data

Secondary language: English
Secondary title: A density improvement method for few-shot object counting
Secondary abstract: Recently, significant progress has been made in the field of few-shot object counting. The current most successful global counters are based on predicting density maps, whose sum estimates the number of objects. However, a major drawback of these methods is that the results are not interpretable, as they do not provide object locations, which is crucial for many applications. In this thesis, we propose the CVDFC method, which uses a diffusion model to enhance the quality of density maps and convert them into precise object location points. The proposed model employs a conditional diffusion process to generate location points, and an object counting module performs non-maximum suppression (NMS) on the generated points, enabling accurate counting and localization of objects in the image. Experimental results showed that the CVDFC method outperforms the reference method LOCA combined with object localization via NMS by 30% in the task of object counting. CVDFC has also proven competitive compared to other methods, demonstrating its effectiveness and practical utility in few-shot object counting.
Secondary keywords: diffusion models;object counting;neural networks;localization points;density map;computer and information science;diploma;
Type (COBISS): Bachelor thesis/paper
Study programme: 1000468
Embargo end date (OpenAIRE): 1970-01-01
Thesis comment: Univ. v Ljubljani, Fak. za računalništvo in informatiko
Pages: 1 spletni vir (1 datoteka PDF (46 str.))
ID: 24985112