izpis_h1_title_alt

Object detection with constellations of keypoints : masters thesis
Rački, Domen (Author), Kristan, Matej (Mentor) More about this mentor... This link opens in a new window, Pock, Thomas (Co-mentor)

URLURL - Presentation file, Visit http://eprints.fri.uni-lj.si/2984/ New window

Abstract
Metode za detekcijo objektov osnovane na značilnicah se za določitev lokacije specifičnega objekta v testni sliki zanašajo na diskriminativno naravo značilnic. Nediskriminativne značilnice v množici detektiranih značilnic se izloča z uporabo podobnostnega pragu. To pomeni da se detektirano značilnico zavrže, če je ta podobna več kot eni značilnici v modelu. V primerih detekcije objektov s ponovljivimi se vzorci se podobnosti prag izkaže kot neučinkovit, saj obravnava večino detektiranih značilnic kot nediskriminativne, t.j., podobnih več kot eni značilnici v modelu. V kontekstu učenja z enim primerom v magistrski nalogi predlagamo konstelacijski model kot dodatek k osnovnim metodam za detekcijo objektov, osnovanih na zančilnicah. Cilj je uporabiti ohranjeno geometrijo med značilnicami kot filter za nediskriminativne značilnice in posledično eliminirati potrebo po podobnostnem pragu. Delovanje predlaganega konstelacijskega modela z empirično in numerično varianco značilnic primerjamo z osnovnim modelom osnovanim na značilnicah. Model evaluiramo na zahtevni bazi katera se sestoji iz logotipov v realnih okoljih. Ugotovimo da je najboljša različica konstelacijskega modela tista z empirično varianco značilnic, saj slednja značilno zmanjša število nediskriminativnih značilnic brez značilnega poslabšanja delovanja algoritma za detekcijo objektov.

Language:English
Keywords:one-shot learning, keypoints, geometry, variance, constellation, object detection, SIFT, GHT, MLESAC, MND, computer science, computer and information science, master's degree
Work type:Master's thesis/paper (mb22)
Tipology:2.09 - Master's Thesis
Organization:FRI - Faculty of computer and information science
Year:2015
Publisher:[D. Rački]
Number of pages:XVIII, 81 str.
UDC:004.932.72(043.2)
COBISS.SI-ID:1536313539 Link is opened in a new window
Views:401
Downloads:251
Metadata:XML RDF-CHPDL DC-XML DC-RDF
 
Average score:(0 votes)
Your score:Voting is allowed only to logged in users.
:
Share: Bookmark and Share

Secondary language

Language:Slovenian
Title:Object detection with constellations of keypoints
Abstract:
Feature-based object detection methods rely on the discriminative nature of features in order to accurately determine the location of a specific object in a test image. From a set of detected features, non-discriminative features are filtered out by means of a similarity threshold, meaning that if a features is very similar to more than one model feature, it is considered to be non-discriminative. However, in cases where an object consists of repeating patterns the similarity threshold proves inefficient since it considers the majority of detected features to be similar to more than one model feature, i.e., non-discriminative. In the context of one-shot learning we propose a constellation model for enhancing basic feature-based object detection methods, with the aim in utilizing the preserved geometry between features to filter out noisy feature matches. This eliminates the need for the similarity threshold. We evaluate the proposed constellation model whit empirically and numerically modelled feature variance and compare it to a baseline feature model. Model evaluation is performed on a challenging real-world dataset, consisting of logotypes in real-world scenarios. We find that the best variation of the constellation model is the model with empirically determined feature variance, which significantly reduces the number of mismatched features, without significantly affecting detection performance.

Keywords:učenje z enim primerom, značilnice, geometrija, varianca, konstelacije, detekcija objektov, SIFT, GHT, MLESAC, MND, računalništvo, računalništvo in informatika, magisteriji

Similar documents

Similar works from RUL:
Similar works from other Slovenian collections:

Comments

Leave comment

You have to log in to leave a comment.

Comments (0)
0 - 0 / 0
 
There are no comments!

Back