An Introduction to Deep Convolutional Neural Nets for Computer Vision

Srinivas, S.; Sarvadevabhatla, Ravi Kiran; Mopuri, Konda Reddy; Prabhu, Nikita; Kruthiventi, Srinivas S. S.; Babu, R. Venkatesh

doi:10.1016/b978-0-12-810408-8.00003-1

Cited by 28 publications

(22 citation statements)

References 24 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We used Python Matplotlib library [32] to generate the scanpath images. CNNs have been shown to be good at detecting local patterns within images [2,54].…”

Section: Generating Scanpath Imagesmentioning

confidence: 99%

Relevance Prediction from Eye-movements Using Semi-interpretable Convolutional Neural Networks

Bhattacharya

Rakshit

Gwizdka

et al. 2020

Proceedings of the 2020 Conference on Human Information Interaction and Retrieval

View full text Add to dashboard Cite

We propose an image-classification method to predict the perceivedrelevance of text documents from eye-movements. An eye-tracking study was conducted where participants read short news articles, and rated them as relevant or irrelevant for answering a trigger question. We encode participants' eye-movement scanpaths as images, and then train a convolutional neural network classifier using these scanpath images. The trained classifier is used to predict participants' perceived-relevance of news articles from the corresponding scanpath images. This method is content-independent, as the classifier does not require knowledge of the screen-content, or the user's information-task. Even with little data, the image classifier can predict perceived-relevance with up to 80% accuracy. When compared to similar eye-tracking studies from the literature, this scanpath image classification method outperforms previously reported metrics by appreciable margins. We also attempt to interpret how the image classifier differentiates between scanpaths on relevant and irrelevant documents.

show abstract

“…We used Python Matplotlib library [32] to generate the scanpath images. CNNs have been shown to be good at detecting local patterns within images [2,54].…”

Section: Generating Scanpath Imagesmentioning

confidence: 99%

Relevance Prediction from Eye-movements Using Semi-interpretable Convolutional Neural Networks

Bhattacharya

Rakshit

Gwizdka

et al. 2020

Proceedings of the 2020 Conference on Human Information Interaction and Retrieval

View full text Add to dashboard Cite

show abstract

“…As it would have been excessively time-consuming to implement and train such a network from scratch, we resorted to a transfer learning solution and built upon an existing and pre-trained model, namely, the VGG16 convolutional neural network (CNN) developed by Simonyan and Zisserman [32]. CNNs are a class of deep neural networks that have been successfully applied for social media image interpretation [33][34][35] as well as in computer vision [36,37] and remote sensing applications [38,39]. The VGG16 network was trained and tested based on ImageNet, a image dataset of over 14 million pictures of about 1000 classes [40].…”

Section: Interpreting the Gsv Imagesmentioning

confidence: 99%

Towards Detecting Building Facades with Graffiti Artwork Based on Street View Images

Novack

Vorbeck

Lorei³

et al. 2020

IJGI

View full text Add to dashboard Cite

As a recognized type of art, graffiti is a cultural asset and an important aspect of a city’s aesthetics. As such, graffiti is associated with social and commercial vibrancy and is known to attract tourists. However, positional uncertainty and incompleteness are current issues of open geo-datasets containing graffiti data. In this paper, we present an approach towards detecting building facades with graffiti artwork based on the automatic interpretation of images from Google Street View (GSV). It starts with the identification of geo-tagged photos of graffiti artwork posted on the photo sharing media Flickr. GSV images are then extracted from the surroundings of these photos and interpreted by a customized, i.e., transfer learned, convolutional neural network. The compass heading of the GSV images classified as containing graffiti artwork and the possible positions of their acquisition are considered for scoring building facades according to their potential of containing the artwork observable in the GSV images. More than 36,000 GSV images and 5000 facades from buildings represented in OpenStreetMap were processed and evaluated. Precision and recall rates were computed for different facade score thresholds. False-positive errors are caused mostly by advertisements and scribblings on the building facades as well as by movable objects containing graffiti artwork and obstructing the facades. However, considering higher scores as threshold for detecting facades containing graffiti leads to the perfect precision rate. Our approach can be applied for identifying previously unmapped graffiti artwork and for assisting map contributors interested in the topic. Furthermore, researchers interested on the spatial correlations between graffiti artwork and socio-economic factors can profit from our open-access code and results.

show abstract

“…A ConvL is a set of learnable filters which actually are three-dimensional matrices, to which a bias vector is attached [1], [2]. The parameters of a ConvL, called hyperparameters, are as follows [2], [3 [5].…”

Section: The Problem Of An Appropriate Numbermentioning

confidence: 99%

“…10. The average performance of the three IRPs by (3) and 4, wherein only three common CNN architectures constitute an argument axis for each of the eight polylines. In the vertical direction, there are not more than two points above the same CNN architecture version.…”

mentioning

confidence: 99%

An attempt of finding an appropriate number of convolutional layers in cnns based on benchmarks of heterogeneous datasets

Romanuke

2018

Electrical, Control and Communication Engineering

View full text Add to dashboard Cite

An attempt of finding an appropriate number of convolutional layers in convolutional neural networks is made. The benchmark datasets are CIFAR-10, NORB and EEACL26, whose diversity and heterogeneousness must serve for a general applicability of a rule presumed to yield that number. The rule is drawn from the best performances of convolutional neural networks built with 2 to 12 convolutional layers. It is not an exact best number of convolutional layers but the result of a short process of trying a few versions of such numbers. For small images (like those in CIFAR-10), the initial number is 4. For datasets that have a few tens of image categories and more, initially setting five to eight convolutional layers is recommended depending on the complexity of the dataset. The fuzziness in the rule is not removable because of the required diversity and heterogeneousness

show abstract

An Introduction to Deep Convolutional Neural Nets for Computer Vision

Cited by 28 publications

References 24 publications

Relevance Prediction from Eye-movements Using Semi-interpretable Convolutional Neural Networks

Relevance Prediction from Eye-movements Using Semi-interpretable Convolutional Neural Networks

Towards Detecting Building Facades with Graffiti Artwork Based on Street View Images

An attempt of finding an appropriate number of convolutional layers in cnns based on benchmarks of heterogeneous datasets

Contact Info

Product

Resources

About