Open AccessLANDMARK-BASED VISUAL PLACE RECOGNITION
Author(s) -
S. L. Gavali,
B. F. Momin
Publication year - 2021
Publication title -
international journal of engineering applied science and technology
Language(s) - English
Resource type - Journals
ISSN - 2455-2143
DOI - 10.33564/ijeast.2021.v05i11.035
Subject(s) - landmark , computer science , artificial intelligence , convolutional neural network , cognitive neuroscience of visual object recognition , pattern recognition (psychology) , object (grammar) , k nearest neighbors algorithm , set (abstract data type) , artificial neural network , computer vision , programming language
— Landmark recognition is one type of problem ofobject recognition that has not been well solved. Theclassical techniques used for object recognition can not beapplied directly due to a large number of landmarks and ahighly imbalanced dataset. This paper presents theapplication of a triplet network for large scale landmarkbased visual place recognition. By fine-tuning pre-trainedconvolutional neural network (CNN) and minimizingtriplet loss, the triplet network can learn appropriatemetrics so that most similar images can be retrievedthrough algorithms for the k-nearest neighbor (KNN). Theperformance of the proposed method is evaluated on adata set for the recognition of real-world landmarks.