Analiza popunjenosti proizvoda na policama temeljena na računalnom vidu

master's thesis

• Title (english): Computer Vision-based Occupancy Analysis in Product Shelf Images
• Author: Franko Šikić
• Mentor: Sven Lončarić (mentor)
• Committee member: Sven Lončarić (predsjednik povjerenstva)
• Committee member: Marko Subašić (član povjerenstva)
• Committee member: Zoran Kalafatić (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; Zagreb
• Defense date and country: 2022-01-27, Croatia
• Scientific / art field, discipline and subdiscipline: TECHNICAL SCIENCES; Computing

Abstract

Popunjenost polica u trgovinama može se analizirati koristeći sustav temeljen na računalnom vidu kako bi se postigli bolji prodajni rezultati ili ostvarilo bolje iskustvo kupovine kod kupaca. Analizu popunjenosti polica moguće je provesti detekcijom proizvoda na policama trgovina, ali i detekcijom praznina na policama. Efikasna detekcija nedostatka proizvoda na policama te brza nadopuna polica s novim proizvodima može povećati prodaju za nekoliko posto, stoga je poseban fokus rada stavljen na detekciju praznina. U radu je navedeno i objašnjeno nekoliko metoda za ostvarivanje detekcije proizvoda i detekcije praznina te je opisan i implementiran potpuno novi pristup u detekciji praznina na policama korištenjem YOLOv5 modela. Naposljetku, prikazani su rezultati detekcije praznina na policama na skupu slika boca snimljenih korištenjem mobilnih uređaja u nekoliko hrvatskih prodajnih lanaca.

Abstract (english)

Occupancy of store shelves can be analysed using a computer vision-based system in order to achieve better sale results or to improve customers' shopping experience. Shelves' occupancy analysis can be carried out by detecting products on store shelves, as well as by detecting void spaces on store shelves. Efficient detection of the missing products on shelves and fast shelf replenishment can boost sales by a few percent, so the focus of this thesis is being put on the detection of void spaces. The thesis lists and describes several approaches to both product detection and void spaces detection, while a novel approach to detection of void spaces on store shelves using YOLOv5 model is described and implemented. Finally, the thesis shows results of the void spaces detection on the dataset which consists of shelf images containing bottles, taken with mobile phones in several Croatian supermarket chains.

• Keywords: računalni vid; analiza slike; analiza popunjenosti polica; detekcija proizvoda; detekcija praznina; konvolucijska neuronska mreža; YOLOv5
• Keywords (english): computer vision; image analysis; shelves' occupancy analysis; product detection; void spaces detection; convolutional neural network; YOLOv5
• Language: croatian
• URN:NBN: urn:nbn:hr:168:879563
• 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 13:55:15