Deep Learning for Visual Indonesian Place Classification with Convolutional Neural Networks

Chowanda, Andry; Sutoyo, Rhio

doi:10.1016/j.procs.2019.08.236

Cited by 5 publications

(6 citation statements)

References 17 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Because it has undergone the process of object recognition, the human visual system is capable of recognising nearly all objects. Object recognition is a technique that recognises the primary properties of an object in order to comprehend it [ 4 ]. This method has two fundamental phases: extraction of features and pattern matching or categorization.…”

Section: Introductionmentioning

confidence: 99%

[Retracted] Environmental and Geographical (EG) Image Classification Using FLIM and CNN Algorithms

Ajay

Nagaraj

Huang

et al. 2022

Contrast Media & Molecular Imaging

View full text Add to dashboard Cite

Intelligent machines have grown in importance in recent years in object recognition in terms of their ability to envision, comprehend, and reach decisions. There are a lot of complicated algorithms that accomplish AI utilities. In addition to their use in the medical industry, these methods of object recognition have a wide range of other fields, most notably industries, in which they can be applied. In contrast to the proposed calculation, the proposed calculation is less complex and more accurate under certain SNR conditions. In the deep nervous tissue fine-tuning discriminator, phantom highlights and binding highlights are separated as sources; modified direct components are used as neuronal activation abilities; and cross entropy is used as unfortunate abilities. Optimized recognition of profound nervous tissue builds profound and periodic nervous tissue for regulatory confirmation of the corresponding signal.

show abstract

Section: Introductionmentioning

confidence: 99%

[Retracted] Environmental and Geographical (EG) Image Classification Using FLIM and CNN Algorithms

Ajay

Nagaraj

Huang

et al. 2022

Contrast Media & Molecular Imaging

View full text Add to dashboard Cite

show abstract

“…The various types of available CNN learning architectures encouraged researcher to know the advantages of each architecture by comparing them on the same dataset. Chowanda et al [12] compared three CNN architectures on a dataset of places/landmark in Indonesia. From many experimental scenarios that have been carried out, the results show that the VGG16 algorithm is superior to VGG19 and GoogleNet with an accuracy value of 92%.…”

Section: Introductionmentioning

confidence: 99%

The effect of Gaussian filter and data preprocessing on the classification of Punakawan puppet images with the convolutional neural network algorithm

Kusrini

Yudianto

Fatta

2022

IJECE

View full text Add to dashboard Cite

<p class="Body"><span>Nowadays, many algorithms are introduced, and some researchers focused their research on the utilization of convolutional neural network (CNN). CNN algorithm is equipped with various learning architectures, enabling researchers to choose the most effective architecture for classification. However, this research suggested that to increase the accuracy of the classification, preprocessing mechanism is another significant factor to be considered too. This study utilized Gaussian filter for preprocessing mechanism and VGG16 for learning architecture. The Gaussian filter was combined with different preprocessing mechanism applied on the selected dataset, and the measurement of the accuracy as the result of the utilization of the VGG16 learning architecture was acquired. The study found that the utilization of using contrast limited adaptive histogram equalization (CLAHE) + red green blue (RGB) + Gaussian filter and thresholding images showed the highest accuracy, 98.75%. Furthermore, another significant finding is that the Gaussian filter was able to increase the accuracy on RGB images, however the accuracy decreased for green channel images. Finally, the use of CLAHE for dataset preprocessing increased the accuracy dealing with the green channel images.</span></p>

show abstract

“…The training of a next-best-view (NBV) planner for visual place recognition (VPR) is fundamentally important for autonomous robot navigation. VPR is typically formulated as a passive single-view image classification problem in visual robot navigation [1,2,3], with the aim of classifying a view image into one of the predefined place classes. However, such passive formulation is ill-posed and its VPR performance is significantly dependent on the presence or absence of landmark-like objects in the view image [4].…”

Section: Introductionmentioning

confidence: 99%

“…1). It has been demonstrated that a CNN classifier that is employed by a standard single-view VPR module [1,2,3] can provide domain-invariant visual cues [11,12,13], such as activations that are invoked by a view image. This motivated us to reformulate the NBV problem to use domain-invariant visual cues instead of a raw view image as the visual input to the NBV planner.…”

Section: Introductionmentioning

confidence: 99%

Active Domain-Invariant Self-Localization Using Ego-Centric and World-Centric Maps

Kurauchi¹,

Yamamoto²,

Yoshida³

2022

Preprint

View full text Add to dashboard Cite

The training of a next-best-view (NBV) planner for visual place recognition (VPR) is a fundamentally important task in autonomous robot navigation, for which a typical approach is the use of visual experiences that are collected in the target domain as training data. However, the collection of a wide variety of visual experiences in everyday navigation is costly and prohibitive for real-time robotic applications. We address this issue by employing a novel domain-invariant NBV planner. A standard VPR subsystem based on a convolutional neural network (CNN) is assumed to be available, and its domain-invariant state recognition ability is proposed to be transferred to train the domain-invariant NBV planner. Specifically, we divide the visual cues that are available from the CNN model into two types: the output layer cue (OLC) and intermediate layer cue (ILC). The OLC is available at the output layer of the CNN model and aims to estimate the state of the robot (e.g., the robot viewpoint) with respect to the world-centric view coordinate system. The ILC is available within the middle layers of the CNN model as a high-level description of the visual content (e.g., a saliency image) with respect to the ego-centric view. In our framework, the ILC and OLC are mapped to a state vector and subsequently used to train a multiview NBV planner via deep reinforcement learning. Experiments using the public NCLT dataset validate the effectiveness of the proposed method.

show abstract

Deep Learning for Visual Indonesian Place Classification with Convolutional Neural Networks

Cited by 5 publications

References 17 publications

[Retracted] Environmental and Geographical (EG) Image Classification Using FLIM and CNN Algorithms

[Retracted] Environmental and Geographical (EG) Image Classification Using FLIM and CNN Algorithms

The effect of Gaussian filter and data preprocessing on the classification of Punakawan puppet images with the convolutional neural network algorithm

Active Domain-Invariant Self-Localization Using Ego-Centric and World-Centric Maps

Contact Info

Product

Resources

About