At the intersection of creative text generation and literary theory, this study explores the role of literary metaphor and its capacity to generate a range of meanings. In this regard, literary metaphor is vital to the development of any particular language. To investigate whether the inclusion of original figurative language improves textual quality, we trained an LSTM-based language model in Afrikaans. The network produces phrases containing compellingly novel figures of speech. Specifically, the emphasis falls on how AI might be utilised as a defamiliarisation technique, which disrupts expected uses of language to augment poetic expression. Providing a literary perspective on text generation, the paper raises thought-provoking questions on aesthetic value, interpretation and evaluation.

Promoted as the first theatrical play written by artificial intelligence, AI: When a Robot Writes a Play (2021) anticipates the post-Anthropocene in both content and form. The machine-generated script problematizes the necessity of human (including dramatic) activities and invites exploration of theatre’s role in the post-anthropocentric condition. AI renders the stage as a transitional site—a potential gateway to other realities.

A dynamic domain in Artificial Intelligence research, Natural Language Generation centres on the automatic generation of realistic text. To help navigate this vast and swiftly developing body of work, the study provides a concise overview of noteworthy stages in the history of text generation. To this end, the paper describes deep learning models for a broad audience, focusing on traditional, convolutional, recurrent and generative adversarial networks, as well as transformer architecture.

Bu çalışmada kontrolsüz ortamda çekilen iki boyutlu imajların birleştirilmesi için tam otomatik bir yöntem sunulmaktadır. Özellikle stereo ve çok açılı kamera uygulamaları, uydu ve sualtı haritalama sistemleri, tıbbi görüntüleme gibi alanlarda ihtiyaç duyulan bu yöntem, görüntülerdeki görüş alanı limitasyonuna bir çözüm sunmaktadır. Çalışma, görüntüler arasında bulunan referans noktaların SURF algoritması ile tespiti, bu noktalar üzerinden RANSAC ve homografi teknikleri ile imajların hizalanması ve Poisson yöntemi ile imajların harmanlanmasına dayanmaktadır. Yapılan nicel ve nitel deneyler, önerilen yöntemin diğer yaklaşımlardan daha kaliteli sonuçlar verdiğini göstermektedir.

The use of artificial intelligence in the legal sector flourished in recent years. This development is often met with excitement and unease. In this critical reflection, we analyse how artificial intelligence functions in modern legal technologies, and what its future implications are for the legal sector and critical legal thinking. We firstly discuss how machine learning and ‘Narrow AI’ are pertinent in this discussion, and how misleading the ‘hype’ on robot lawyers is. We then show how legal technologies are currently utilized, and the potential ways to map the modern legal technology landscape. Finally, we examine the potential effects of AI and legal technologies on legal decision-making, as complex algorithms open up the potential to disarrange or obscure critical analysis.

This study presents a robust, low-cost hand motion recognition approach designed to run on low-end computer systems. Our method detects and tracks hand region using real-time images obtained from a low-resolution camera (i.e. webcam) and is not depended on any training or calibration and is not required any special camera apparatus or selectors. The proposed system involves several image processing techniques such as background subtraction, face detection, skin colour detection and template matching. The experimental results show promising performance under various conditions. The method has a wide range of applications where more natural ways of interaction required, such as virtual reality applications, assistive technologies and simulation.

Recent years have seen a proliferation of attention mechanisms and the rise of Transformers in Natural Language Generation (NLG). Previously, state-of-the-art NLG architectures such as RNN and LSTM ran into vanishing gradient problems; as sentences grew larger, distance between positions remained linear, and sequential computation hindered parallelization since sentences were processed word by word. Transformers usher in a new era. In this paper, we explore three major Transformer-based models, namely GPT, BERT, and XLNet, that carry significant implications for the field. NLG is a burgeoning area that is now bolstered with rapid developments in attention mechanisms. From poetry generation to summarization, text generation derives benefit as Transformer-based language models achieve groundbreaking results.

This paper proposes a generative language model called AfriKI. Our approach is based on an LSTM architecture trained on a small corpus of contemporary fiction. With the aim of promoting human creativity, we use the model as an authoring tool to explore machine-in-the-loop Afrikaans poetry generation. To our knowledge, this is the first study to attempt creative text generation in Afrikaans.

This study reviews the approaches that are based on heuristic methods and include processed GPS data in their solutions for vehicle routing problems. Vehicle routing problems are challenging given their constraints and complexity. Due to the nature of vehicle routing, the appropriate solution must be found within a reasonable timeframe. However, it is not possible to scan the entire solution space and find the optimum solution within the ideal timeframe when the problem's complexity increases because of the number of points and the constraints. In order to solve these kinds of situations, the aim is finding the optimum solution, or the optimum solution as close as possible. This situation reveals some scenarios where heuristic and meta-heuristic algorithms are used as solution algorithms to routing problems. GPS data is included in the solution algorithms to increase the performance of heuristic algorithms and routing solutions. As a result of the processed data, environmental factors such as congestion points, average speed on the route and traffic density according to hours are also taken into account. In this way, more consistent solutions are developed for real-life applications.

Solid waste management, which includes collection, transportation, disposal and recycling, is a critical environmental issue increasing in severity due to industrialization, urban population growth and consumption. In this study, we present a vehicle routing approach based on the ant colony optimization algorithm. We work on the real-world data of Maltepe Municipality, Istanbul, Turkey. The experimental results show that we estimate the shortest route path with 13% efficiency compared to the existing route. We believe this would play an important role in solid waste management, help reduce costs as well as provide fast and efficient waste collection for local municipalities.

With digital transformation and progress in science, developing process capabilities with limited resources has gained immense importance for higher education institutions. To provide the highest quality education services and manage the information and knowledge management processes, appropriate strategies, and related key performance management factors must be defined efficiently. In the literature, numerous models and tools have been proposed to measure knowledge management capabilities (Mohapatra et al., 2016; Imran et al., 2017; Dayan et al., 2017; Trivella et al., 2015). This chapter aims to develop a better understanding of the relationship between knowledge management processes and propose an assessment model of knowledge management system for higher education institutions to measure knowledge resources and their capabilities. To do so, the knowledge management system is identified by four main attributes (People, Processes, Technology and Culture) and each attribute is comprehensively analysed in a dedicated section.

Anticipating the rise in Artificial Intelligence’s ability to produce original works of literature, this study suggests that literariness, or that which constitutes a text as literary, is understudied in relation to text generation. From a computational perspective, literature is particularly challenging because it typically employs figurative and ambiguous language. Literary expertise would be beneficial to understanding how meaning and emotion are conveyed in this art form but is often overlooked. We propose placing experts from two dissimilar disciplines – machine learning and literary studies – in conversation to improve the quality of AI writing. Concentrating on evaluation as a vital stage in the text generation process, the study demonstrates that benefit could be derived from literary theoretical perspectives. This knowledge would improve algorithm design and enable a deeper understanding of how AI learns and generates. This article appears in the special track on AI and Society.

With the increasing amount of data produced and collected, the use of artificial intelligence technologies has become inevitable. By using deep learning techniques from these technologies, high performance can be achieved in tasks such as classification and face analysis in the fields of image processing and computer vision. In this study, Convolutional Neural Networks (CNN), one of the deep learning algorithms, was used. The model created with this algorithm was trained with facial images and gender prediction was made. As a result of the experiments, 93.71% success rate was achieved on the VGGFace2 data set and 85.52% success rate on the Adience data set. The aim of the study is to classify low-resolution images with high accuracy.

Makerspaces are becoming increasingly important as a new approach to enable individuals to experience new technology applications and increase creativity in universities and other educational institutions. Such learning implementations offer many advantages like learning-by-doing and applying theoretical knowledge in engineering education into practical skills in an interdisciplinary environment. Despite the advantages, makerspaces still lack integration into the curriculum of engineering schools. For a quality engineering education, establishing maker workshops where students can experiment, design and practice as well as feel encouraged to open-ended development projects are required in addition to standard theoretical and laboratory applications. In this study, we present student views on the learning opportunities, challenges and contributions of the makerspace environment. In this context, IHA Makerspace established within the Faculty of Technology, Marmara University, Turkey, aims to provide high-level engineering experience to students by designing and prototyping effectively in a multidisciplinary development environment.

Changes in lighting conditions are an important factor for facial recognition applications. The algorithms used in these applications have various approaches and are directly affected by environments under difficult lighting settings. In this study, we investigate two appearance-based local and global approaches, namely Principal Component Analysis and Local Binary Patterns algorithms, examine important studies on these algorithms and compare their facial recognition performances on images from the Extended Yale Face Database B. Experiments show that the LBP method provides better results under varying lighting conditions.

A face image contains geometric cues in the form of configurational information and contours that can be used to estimate 3D face shape. While it is clear that 3D reconstruction from 2D points is highly ambiguous if no further constraints are enforced, one might expect that the face-space constraint solves this problem. We show that this is not the case and that geometric information is an ambiguous cue. There are two sources for this ambiguity. The first is that, within the space of 3D face shapes, there are flexibility modes that remain when some parts of the face are fixed. The second occurs only under perspective projection and is a result of perspective transformation as camera distance varies. Two different faces, when viewed at different distances, can give rise to the same 2D geometry. To demonstrate these ambiguities, we develop new algorithms for fitting a 3D morphable model to 2D landmarks or contours under either orthographic or perspective projection and show how to compute flexibility modes for both cases. We show that both fitting problems can be posed as a separable nonlinear least squares problem and solved efficiently. We demonstrate both quantitatively and qualitatively that the ambiguity is present in reconstructions from geometric information alone but also in reconstructions from a state-of-the-art CNN-based method.

We generalise Spatial Transformer Networks (STN) by replacing the parametric transformation of a fixed, regular sampling grid with a deformable, statistical shape model which is itself learnt. We call this a Statistical Transformer Network (StaTN). By training a network containing a StaTN end-to-end for a particular task, the network learns the optimal nonrigid alignment of the input data for the task. Moreover, the statistical shape model is learnt with no direct supervision (such as landmarks) and can be reused for other tasks. Besides training for a specific task, we also show that a StaTN can learn a shape model using generic loss functions. This includes a loss inspired by the minimum description length principle in which an appearance model is also learnt from scratch. In this configuration, our model learns an active appearance model and a means to fit the model from scratch with no supervision at all, even identity labels.

Aishwarya Agrawal Amit Agrawal Antonio Agudo Zeynep Akata Shuichi Akizuki Karteek Alahari Xavier Alameda-Pineda Andrea Albarelli Michel Antunes Pablo Arbelaez Freddie Åström Vassilis Athitsos Mathieu Aubry Chloé-Agathe Azencott Hossein Azizpour Andrew Bagdanov Xiang Bai Guha Balakrishnan Adrian Barbu Jonathan Barron Anil Bas Dhruv Batra Maximilian Baust Jean Charles Bazin Loris Bazzani Fabian Benitez-Quiroz Martin Benning Ryad Benosman Binod Bhattarai Silvia Biasotti Pierre Biasutti Hakan Bilen Thomas Bishop Lubomir Bourdev Eric Brachmann Michael Brown Andres Bruhn Antoni Buades Aurélie Bugeau Tien Bui W. Burgard Darius Burschka Yohann Cabon Carlos Castillo Luka Cehovin Ayan Chakrabarti Rudrasis Chakraborty Antoni Chan Manmohan Chandraker Wei Chao Nicolas Charon Guillaume Charpiat Ankur Chattopadhyay Jun-Cheng Chen Jiansheng Chen Xiaowu Chen Qifeng Chen Xinlei Chen Tat-Jun Chin Wen-Sheng Chu Albert Chung Ramazan Gokberk CINBIS Toby Collins Nicolo Colombo Oliver Cossairt Camille Couprie Nathan Crombez Gabriela Csurka Martin Danelljan Mohamed Daoudi Andrew Davison fernando De la Torre Cesar De Souza Bartolomeo Della Corte Joachim Denzler Chaitanya Desai Santosh Divvala Puneet Dokania Jian Dong Matthijs Douze Birgitta Dresp-Langley Enrique Dunn Anjan Dutta Segio Escalera Francisco Escolano Virginia Estellers Georgios Evangelidis Xiaochuan Fan A. Fitzgibbon Alessandro Foi Denis Fortun David Fouhey Uwe Franke Oren Freifeld Mario Fritz Zhenyong Fu Yasutaka Furukawa Yasutaka Furukawa Donald G. Dansereau Juergen Gall Xinbo Gao Simone Gasparini Efstratios Gavves D. Geiger Bernard Ghanem Guy Gilboa Ioannis Gkioulekas Georgia Gkioxari Daniel Glasner Alvina Goh Bastian Goldluecke Boqing Gong Stephen Gould Kristen Grauman Rafael Grompone von Gioi Jose J. Guerrero Matthieu Guillaumin Jean-Yves Guillemaut Yulan Guo Yanwen Guo Xiaohu Guo Yuhong Guo Saurabh Gupta

In this paper, we show how a 3D Morphable Model (i.e. a statistical model of the 3D shape of a class of objects such as faces) can be used to spatially transform input data as a module (a 3DMM-STN) within a convolutional neural network. This is an extension of the original spatial transformer network in that we are able to interpret and normalise 3D pose changes and self-occlusions. The trained localisation part of the network is independently useful since it learns to fit a 3D morphable model to a single image. We show that the localiser can be trained using only simple geometric loss functions on a relatively small dataset yet is able to perform robust normalisation on highly uncontrolled images including occlusion, self-occlusion and large pose changes.

In this paper we explore the problem of fitting a 3D morphable model to single face images using only sparse geometric features (edges and landmark points). Previous approaches to this problem are based on nonlinear optimisation of an edge-derived cost that can be viewed as forming soft correspondences between model and image edges. We propose a novel approach, that explicitly computes hard correspondences. The resulting objective function is non-convex but we show that a good initialisation can be obtained efficiently using alternating linear least squares in a manner similar to the iterated closest point algorithm. We present experimental results on both synthetic and real images and show that our approach outperforms methods that use soft correspondence and other recent methods that rely solely on geometric features.