A Survey on Deep Learning

Pouyanfar, Samira; Sadiq, Saad; Yan, Yilin; Tian, Haiman; Tao, Yudong; Reyes, Maria Presa; Shyu, Mei‐Ling; Chen, Shu‐Ching; Iyengar, S. Sitharama

doi:10.1145/3234150

Cited by 1,098 publications

(296 citation statements)

References 115 publications

Supporting

Mentioning

291

Contrasting

Unclassified

Order By: Relevance

“…and with a broad range of applications [49]. Within this domain, a popular approach to try and make those systems explainable is to produce "saliency maps" (also called "heat-maps") that highlight which pixels were most important for the image classi cation algorithm.…”

Section: Introductionmentioning

confidence: 99%

Evaluating saliency map explanations for convolutional neural networks

Alqaraawi

Schuessler

Weiß

et al. 2020

Proceedings of the 25th International Conference on Intelligent User Interfaces

137

View full text Add to dashboard Cite

Convolutional neural networks (CNNs) o er great machine learning performance over a range of applications, but their operation is hard to interpret, even for experts. Various explanation algorithms have been proposed to address this issue, yet limited research e ort has been reported concerning their user evaluation. In this paper, we report on an online between-group user study designed to evaluate the performance of "saliency maps" -a popular explanation algorithm for image classi cation applications of CNNs. Our results indicate that saliency maps produced by the LRP algorithm helped participants to learn about some speci c image features the system is sensitive to. However, the maps seem to provide very limited help for participants to anticipate the network's output for new images. Drawing on our ndings, we highlight implications for design and further research on explainable AI. In particular, we argue the HCI and AI communities should look beyond instance-level explanations.

show abstract

Section: Introductionmentioning

confidence: 99%

Evaluating saliency map explanations for convolutional neural networks

Alqaraawi

Schuessler

Weiß

et al. 2020

Proceedings of the 25th International Conference on Intelligent User Interfaces

137

View full text Add to dashboard Cite

show abstract

“…We select two common and widely research applications of the field [11], namely image classification and object detection. In both case studies, we identify fault-revealing MRs and robustness boundaries against configurable image transformations.…”

Section: Case Studiesmentioning

confidence: 99%

“…Third, we explore the benefits of Tetraband on two case studies coming from image analysis, namely image classification and object recognition. Both of these case studies are relevant subsystems in a wide number of applications, such as autonomous cars, robot navigation, or industrial automation [10,11]. For these applications, high-quality standards are essential and rigorous testing is a requirement.…”

Section: Introductionmentioning

confidence: 99%

Adaptive metamorphic testing with contextual bandits

Spieker

Gotlieb

2020

Journal of Systems and Software

View full text Add to dashboard Cite

Metamorphic Testing is a software testing paradigm which aims at using necessary properties of a systemunder-test, called metamorphic relations, to either check its expected outputs, or to generate new test cases. Metamorphic Testing has been successful to test programs for which a full oracle is not available or to test programs for which there are uncertainties on expected outputs such as learning systems. In this article, we propose Adaptive Metamorphic Testing as a generalization of a simple yet powerful reinforcement learning technique, namely contextual bandits, to select one of the multiple metamorphic relations available for a program. By using contextual bandits, Adaptive Metamorphic Testing learns which metamorphic relations are likely to transform a source test case, such that it has higher chance to discover faults. We present experimental results over two major case studies in machine learning, namely image classification and object detection, and identify weaknesses and robustness boundaries. Adaptive Metamorphic Testing efficiently identifies weaknesses of the tested systems in context of the source test case.

show abstract

“…As datasets have become larger, there has been a growing interest in deploying deep learning-based (DL) approaches to the problem of road and asphalt damage detection. It well known [13] that DL algorithms have demonstrated impressive results in a variety of computer vision-related fields and are behind the autonomous driving revolution. In this sense, several efforts have been conducted to leverage these capabilities in the problem we are tackling, with some very impressive results for certain classes of asphalt damages such as cracks [14,15].…”

Section: Related Workmentioning

confidence: 99%

Road Damage Detection Acquisition System Based on Deep Neural Networks for Physical Asset Management

Angulo¹,

Vega-Fernández²,

Aguilar-Lobo³

et al. 2019

Lecture Notes in Computer Science

View full text Add to dashboard Cite

Research on damage detection of road surfaces using image processing techniques has been actively conducted, achieving considerably high detection accuracies. So far, most studies focused only on the detection of the presence or absence of damages; however, in real-world scenarios, road managers from need to clearly understand the type of damage and its extent in order to take effective action in advance or to allocate the necessary resources. Moreover, currently there are few uniform and openly available road damage datasets available, leading to a lack of a common benchmark for road damage detection. Such dataset could be used for a great variety of applications; herein, it is intended to serve as the acquisition component of a physical asset management tool which can aid governments agencies for planning purposes, or by infrastructure maintenance companies, so they can implement predictive maintenance procedures.In this paper, we make two contributions to address these issues. First, we present a large-scale road damage dataset, which includes a more balanced and representative set of damages not present in previous studies. This dataset is composed of 18,0345 road damage images captured with a smartphone installed on a car, with 45,435 instances road surface damages (linear, lateral and alligator cracks, potholes, and various types of painting blurs). In order to generate this dataset, we obtained images from several public datasets and augmented it with crowdsourced images, which where manually annotated for further processing. The images were captured under a variety of weather and illumination conditions. In each image, we annotated the bounding box representing the location and type of damage and its extent. Second, we trained different types generic object detection methods, both traditional (an LBP-cascaded classifier) and deep learning-based, specifically, MobileNet and RetinaNet, which are amenable for embedded and mobile and implementations with an acceptable performance for many applications. We compared the accuracy and inference time of all these models with others in the state of the art, achieving higher accuracies in all the eight classes present in the dataset introduced by researchers at the University of Tokyo, and in other related works, with a lower inference time.

show abstract

A Survey on Deep Learning

Cited by 1,098 publications

References 115 publications

Evaluating saliency map explanations for convolutional neural networks

Evaluating saliency map explanations for convolutional neural networks

Adaptive metamorphic testing with contextual bandits

Road Damage Detection Acquisition System Based on Deep Neural Networks for Physical Asset Management

Contact Info

Product

Resources

About