Konvolucijski modeli za klasifikaciju videa

master's thesis

• Title (english): Convolutional models for video classification
• Author: Antonio Borac
• Mentor: Siniša Šegvić (mentor)
• Committee member: Siniša Šegvić (predsjednik povjerenstva)
• Committee member: Slobodan Ribarić (član povjerenstva)
• Committee member: Vlado Glavinić (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; Zagreb
• Defense date and country: 2020-07-08, Croatia
• Scientific / art field, discipline and subdiscipline: TECHNICAL SCIENCES; Computing

Abstract

Klasifikacija videa neriješen je problem računalnog videa s mnogim zanimljivim primjenama. U posljednje vrijeme najbolji rezultati u tom području postižu se pristupima temeljenim na dubokim konvolucijskim modelima. Za ovaj rad posebno su zanimljivi nadzirani pristupi gdje je svaki video označen razredom radnje koju video predstavlja. U okviru rada proučio sam postojeće pristupe klasifikaciji videa. Objasnio sam teorijsku podlogu bitnu za razumijevanje eksperimenata koji se provode. Izgradio sam dvije konkretne arhitekture za klasifikaciju videa. Izrađena implementacija modela temelji se na radnom okviru Keras. Opisao sam i arhitekture na kojima se arhitekture od interesa baziraju. Prikazao sam i ocijenio ostvarene rezultate i predložio pravce za budući razvoj.

Abstract (english)

Video classification is an unsolved problem of computer vision with many interesting applications. Lately, best results in that area are obtained using approaches based on deep convolutional models. This thesis considers supervised approaches where every video is labeled with class of action the video represents. I examined existing approaches to video classification which were interesting for this paper. I explained the theory necessary for understanding of the experiments. Model implementation is based on Keras framework. I have built two distinct architectures for video classification. Additionally, I have explained the architectures on which the architectures of interest are based. I reported the results and compared them with results from literature and listed possible improvements and directions for future work.

• Keywords: duboko učenje; klasifikacija videa; konvolucijski modeli za klasifikaciju videa; Keras; UCF-101; SlowFast; sažimanje značajki
• Keywords (english): deep learning; video classification; convolutional models for video classification; Keras; UCF-101; SlowFast; feature-pooling
• Language: croatian
• URN:NBN: urn:nbn:hr:168:751974
• Study programme: Title: Computing; Study programme type: university; Study level: graduate; Academic / professional title: magistar/magistra inženjer/inženjerka računarstva (magistar/magistra inženjer/inženjerka računarstva)
• Type of resource: Text
• File origin: Born digital
• Access conditions: Closed access
• Created on: 2023-01-13 11:49:57