Permanent URI for this collection
Browse
Recent Submissions
Publication Conference program and online proceedings of the Irish Machine Vision and Image Processing Conference 2023(University of Galway, 2023-08-30) Corcoran, Peter; Schukat, Michael; Murray, Niall; Farooq, Muhammad Ali; Javidnia, HoseinThe Irish Machine Vision and Image Processing Conference (IMVIP) Conference is the annual research conference of the Irish Pattern Recognition and Classification Society. The chief objective of the society is the advancement of research and study of pattern recognition, classification and associated fields and the applications of the outcomes of research. IPRCS is a member of the International Association for Pattern Recognition (IAPR) and the International Federation of Classification Societies.Publication Comparison of three buck topologies for wide output voltage applications(IEEE, 2023-05-31) Anderson, Oisín; Barry, Brendan; Hogan, Diarmuid; Duffy, Maeve; Irish Research CouncilThis paper investigates the suitability of three step-down dc-dc converter topologies as the final conversion stage in a wide output voltage modular ac-dc power supply. Single-phase, two-phase and three-level buck converters are evaluated over a wide range of outputs using analytical models and physical measurements. The converters' performance is evaluated at all operating points using statistical analysis of the converter component losses produced to assess their suitability for wide output voltage applications. The dynamic performance of the converters is also evaluated to determine their stability for on-the-fly variations in output voltage and load. The analysis finds that the three-level converter is more efficient across the full output range, with lower component loss variability compared to the one-phase and two-phase buck converters. However, it suffers from poor dynamic performance with high output deviations and slow response times. The analysis was verified using three prototype converters designed for 200 W, 15 V to 28 V output.Publication Iris recognition on consumer devices - challenges and progress(IEEE, 2015-11-11) Thavalengal, Shejin; Corcoran, PeterThis article outlines various technical, social and ethical challenges in implementing and widely adopting iris recognition technology on consumer devices such as smartphone or tablets. Acquisition of sufficient quality iris images using devices is noted to be the main challenge in implementing this technology. Current progress in this field is reviewed. A smartphone form factor camera is presented to be used as a front-facing camera. This device is modified to capture near infra-red iris images along with general purpose visible wavelength images. Analyses shows that such a device with improved optics and sensor could be used for implementing iris recognition in next generation hand held devices. The social impact of wider adoption of this technology is discussed. Iris pattern obfuscation is presented to address various security and privacy concerns which may arise when iris recognition will be a part of our daily life.Publication High-accuracy facial depth models derived from 3D synthetic data(Institute of Electrical and Electronics Engineers (IEEE), 2020-08-31) Khan, Faisal; Basak, Shubhajit; Javidnia, Hossein; Schukat, Michael; Corcoran, PeterIn this paper, we explore how synthetically generated 3D face models can be used to construct a high-accuracy ground truth for depth. This allows us to train the Convolutional Neural Networks (CNN) to solve facial depth estimation problems. These models provide sophisticated controls over image variations including pose, illumination, facial expressions and camera position. 2D training samples can be rendered from these models, typically in RGB format, together with depth information. Using synthetic facial animations, a dynamic facial expression or facial action data can be rendered for a sequence of image frames together with ground truth depth and additional metadata such as head pose, light direction, etc. The synthetic data is used to train a CNN-based facial depth estimation system which is validated on both synthetic and real images. Potential fields of application include 3D reconstruction, driver monitoring systems, robotic vision systems, and advanced scene understanding.Publication Generative Augmented Dataset and Annotation Frameworks for Artificial Intelligence (GADAFAI)(Institute of Electrical and Electronics Engineers (IEEE), 2020-08-31) Corcoran, Peter; Javidnia, Hossein; Lemley, Joseph E.; Varkarakis, ViktorRecent Advances in Artificial Intelligence (AI), particularly in the field of compute vision, have been driven by the availability of large public datasets. However, as AI begins to move into embedded devices there will be a growing need for tools to acquire and re-acquire datasets from specific sensing systems to train new device models. In this paper, a roadmap in introduced for a data-acquisition framework that can build the large synthetic datasets required to train AI systems from small seed datasets. A key element to justify such a framework is the validation of the generated dataset and example results are shown from preliminary work on biometric (facial) datasets.Publication A novel machine learning based method for deepfake video detection in social media(Institute of Electrical and Electronics Engineers (IEEE), 2021-05-12) Mitra, Alakananda; Mohanty, Saraju P.; Corcoran, Peter; Kougianos, EliasWith the advent of deepfake videos, video forgery has become a serious threat. Videos in social media are the most common and serious targets. There are some existing works for detecting deepfake videos but very few attempts have been made for videos in social media. This paper presents a neural network based method to detect fake videos. A model, consisting of a convolutional neural network (CNN) and a classifier network is proposed. Three different structures, XceptionNet, InceptionV3 and Resnet50 have been considered as the CNN modules and a comparative study has been made. Xception Net has been chosen in the proposed model and paired with the proposed classifier for classification. We used the FaceForensics++ dataset to reach the best model. Our model integrated in the algorithm detects compressed videos in social media.Publication Accurate 2D facial depth models derived from a 3D synthetic dataset(Institute of Electrical and Electronics Engineers (IEEE), 2021-05-13) Khan, Faisal; Basak, Shubhajit; Corcoran, PeterAs Consumer Technologies (CT) seeks to engage and interact more closely with the end-user it becomes important to observe and analyze a user’s interaction with CT devices and associated services. One of the most useful modes for monitoring a user is to analyze a real-time video stream of their face. Facial expressions, movements and biometrics all provide important information, but obtaining a calibrated input with 3D accuracy from a single camera requires accurate knowledge of the facial depth and distance of different features from the camera. In this paper, a method is proposed to generate synthetic high-accuracy human facial depth from synthetic 3D face models. The generated synthetic human facial dataset is then used in Convolutional Neural Networks (CNN’s) for monocular depth facial estimation and the results of the experiments are presented.Publication Re-training StyleGAN-A first step towards building large, scalable synthetic facial datasets(Institute of Electrical and Electronics Engineers (IEEE), 2020-08-31) Varkarakis, Viktor; Bazrafkan, Shabab; Corcoran, Peter; Science Foundation Ireland; FotoNation LtdStyleGAN is a state-of-art generative adversarial network architecture that generates random 2D high-quality synthetic facial data samples. In this paper we recap the StyleGAN architecture and training methodology and present our experiences of retraining it on a number of alternative public datasets. Practical issues and challenges arising from the retraining process are discussed. Tests and validation results are presented and a comparative analysis of several different re-trained StyleGAN weightings is provided. The role of this tool in building large, scalable datasets of synthetic facial data is also discussed.Publication Infrared imaging for human thermography and breast tumor classification using thermal images(Institute of Electrical and Electronics Engineers (IEEE), 2020-08-31) Farooq, Muhammad Ali; Corcoran, Peter; Horizon 2020; Enterprise Ireland International Research FundHuman thermography is considered to be an integral medical diagnostic tool for detecting heat patterns and measuring quantitative temperature data of the human body. It can be used in conjunction with other medical diagnostic procedures for getting comprehensive medication results. In the proposed study we have highlighted the significance of Infrared Thermography (IRT) and the role of machine learning in thermal medical image analysis for human health monitoring and various disease diagnosis in preliminary stages. The first part of the proposed study provides comprehensive information about the application of IRT in the diagnosis of various diseases such as skin and breast cancer detection in preliminary stages, dry eye syndromes, and ocular issues, liver disease, diabetes diagnosis and last but not least the novel COVID-19 virus. Whereas in the second phase we have proposed an autonomous breast tumor classification system using thermal breast images by employing state of the art Convolution Neural Network (CNN). The system achieves the overall accuracy of 80% and recall rate of 83.33%.Publication Generating thermal image data samples using 3D facial modelling techniques and deep learning methodologies(Institute of Electrical and Electronics Engineers (IEEE), 2020-05-26) Farooq, Muhammad Ali; Corcoran, Peter; Horizon 2020Methods for generating synthetic data have become of increasing importance to build large datasets required for Convolution Neural Networks (CNN) based deep learning techniques for a wide range of computer vision applications. In this work, we extend existing methodologies to show how 2D thermal facial data can be mapped to provide 3D facial models. For the proposed research work we have used tufts datasets for generating 3D varying face poses by using a single frontal face pose. The system works by refining the existing image quality by performing fusion based image preprocessing operations. The refined outputs have better contrast adjustments, decreased noise level and higher exposedness of the dark regions. It makes the facial landmarks and temperature patterns on the human face more discernible and visible when compared to original raw data. Different image quality metrics are used to compare the refined version of images with original images. In the next phase of the proposed study, the refined version of images is used to create 3D facial geometry structures by using Convolution Neural Networks (CNN). The generated outputs are then imported in blender software to finally extract the 3D thermal facial outputs of both males and females. The same technique is also used on our thermal face data acquired using prototype thermal camera (developed under Heliaus EU project) in an indoor lab environment which is then used for generating synthetic 3D face data along with varying yaw face angles and lastly facial depth map is generated.Publication Proof-of-concept techniques for generating synthetic thermal facial data for training of deep learning models(National University of Ireland Galway, 2021-01-10) Farooq, Muhammad Ali; Corcoran, Peter; Horizon 2020; Enterprise Ireland International Research FundThermal imaging has played a dynamic role in the diversified field of consumer technology applications. To build artificially intelligent thermal imaging systems, large scale thermal datasets are required for successful convergence of complex deep learning models. In this study, we have highlighted various techniques for generating large scale synthetic facial thermal data using both public and locally gathered datasets. It includes data augmentation, synthetic data generation using StyleGAN network, and 2D to 3D image reconstruction using deep learning architectures. Training and validation accuracy of Wide ResNet CNN for binary gender recognition task is improved by 4.6% and 4.4% using original and newly generated synthetic data with an overall test accuracy of 83.33%.Publication Microwave bone imaging: experimental evaluation of calcaneus bone phantom and imaging prototype(Institute of Electrical and Electronics Engineers, 2020-12-14) Amin, Bilal; Sheridan, Colin; Kelly, Daniel; O'Halloran, Martin; Elahi, Muhammad Adnan; Horizon 2020 Framework Programme; Horizon 2020Microwave imaging (MWI) can be used as an alternate imaging modality for monitoring bone health. Evaluation and characterization of MWI prototype is a precursor step before in vivo investigation of bone dielectric properties. This paper presents experimental evaluation of a novel two layered simplified cylindrical shaped 3D printed human calcaneus bone phantom along with corresponding MWI prototype designed to image the bone phantom. The shape of the calcaneus bone was approximated with a cylinder. The external and internal layers represent cortical bone and trabecular bone respectively. Each layer of the phantom was filled with respective liquid tissue mimicking mixture (TMM). A MWI prototype was designed having six microstrip antennas in order to hold calcaneus bone phantom. The bone phantom was placed in the imaging prototype and scattered signals were measured at each antenna. Moreover, the performance of the system was explored by examining microwave measurement sensitivity. Based on the measured scattered signals the map of dielectric properties will be constructed by employing MWI algorithm and will be communicated in our future work. This two layered 3D printed human calcaneus bone phantom and imaging prototype can be used as a valuable test platform for pre-clinical assessment of calcaneus bone imaging for monitoring osteoporosis.Publication In-camera person-indexing of digital images(Institute of Electrical and Electronics Engineers, 2006-01-07) Costache, G.; Mulryan, R.; Steinberg, E.; Corcoran, PeterA prototype implementation of an automatic in-camera cataloging tool is presented. An external infrastructure to store and analyze images and support the in-camera cataloging tool is also described.Publication Deep learning for facial expression recognition: A step closer to a smartphone that knows your moods(Institute of Electrical and Electronics Engineers, 2017-01-08) Bazrafkan, Shabab; Nedelcu, Tudor; Filipczuk, Pawel; Corcoran, Peter; Science Foundation Ireland; FotoNation Limited; Irish Research CouncilBy growing the capacity and processing power of the handheld devices nowadays, a wide range of capabilities can be implemented in these devices to make them more intelligent and user friendly. Determining the mood of the user can be used in order to provide suitable reactions from the device in different conditions. One of the most studied ways of mood detection is by using facial expressions, which is still one of the challenging fields in pattern recognition and machine learning science.Deep Neural Networks (DNN) have been widely used in order to overcome the difficulties in facial expression classification. In this paper it is shown that the classification accuracy is significantly lower when the network is trained with one database and tested with a different database. A solution for obtaining a general and robust network is given as well.Publication Real-time automotive street-scene mapping through fusion of improved stereo depth and fast feature detection algorithms(Institute of Electrical and Electronics Engineers, 2017-01-08) Javidnia, Hossein; Corcoran, Peter; Science Foundation Ireland; FotoNation LimitedThe real-time tracking of street scenes as a vehicle is driving is a key enabling technology for autonomous vehicles. In this work we provide the basis for such a system through combining an improved advanced random walk with restart technique for stereo depth determination with fast, robust feature detection. The enables tracking and mapping of a wide range of scene structures which can be readily resolved into individual objects and scene elements. Thus it is practical to identify moving objects such as vehicles, pedestrians and fixed objects and structures such as buildings, trees and roadside kerb.Publication Gaze Visual - A graphical software tool for performance evaluation of eye gaze estimation systems(Institute of Electrical and Electronics Engineers, 2018-08-15) Kar, Anuradha; Corcoran, Peter; Science Foundation Ireland; FotoNation LimitedThe concept of an open source software developed for all round performance evaluation of gaze tracking systems is presented. The capabilities of this software towards quantitative, statistical and visual analysis of gaze data are discussed. Potential utilities of this software are towards understanding a gaze tracker's behavior, gaze data quality, and improving usability of gaze based applications that are currently very popular in augmented/virtual reality, gaming and multimedia domains.Publication A review of resolution losses for AR/VR foveated imaging applications(Institute of Electrical and Electronics Engineers, 2018-08-15) Cognard, Timothee E.; Goncharov, Alexander; Devaney, Nicholas; Dainty, Chris; Corcoran, Peter; Science Foundation Ireland; FotoNation LimitedFoveated imaging is of great interest for Augmented and Virtual Reality applications. The resolution losses off-axis simulated in foveated imaging are modelled using cone density on the retina. This article reviews the other factors limiting the resolution off-axis in AR/VR, in particular the impact of the eye lens. Several off-axis resolution simulations are proposed and compared in order to provide some theoretical compression ratios for 4K and 8K display systems. A model taking into account both the cone density across the retina and the optical performance of the eye lens is proposed and evaluated. The variability and challenges of modelling the human eye resolution losses are also discussed, in particular in the case of age-dependence.Publication Synthesizing game audio using deep neural networks(Institute of Electrical and Electronics Engineers, 2018-08-15) McDonagh, Aoife; Lemley, Joseph; Cassidy, Ryan; Corcoran, Peter; Science Foundation Ireland; FotoNation Limited; Irish Research CouncilHigh quality audio plays an important role in gaming, contributing to player immersion during gameplay. Creating audio content which matches overall theme and aesthetic is essential, such that players can become fully engrossed in a game environment. Sound effects must also fit well with visual elements of a game so as not to break player immersion. Producing suitable, unique sound effects requires the use of a wide range of audio processing techniques. In this paper, we examine a method of generating in-game audio using Generative Adversarial Networks, and compare this to traditional methods of synthesizing and augmenting audio.Publication Deep learning for hand segmentation in complex backgrounds(Institute of Electrical and Electronics Engineers, 2018-01-02) Ungureanu, Adrian-Stefan; Bazrafkan, Shabab; Corcoran, PeterThis paper presents a Deep Learning segmentation approach for hand segmentation in gray level images with cluttered backgrounds where standard techniques cannot be used. Two networks were trained with a database of hand images derived from widely used palmprint image databases, Hong Kong Polytechnic University (HKPU) and Chinese Academy of Science (CASIA). The image dataset is augmented with complex patterns and used to train and test the Neural Networks, providing promising results.Publication The application of deep learning on depth from multi-array camera(Institute of Electrical and Electronics Engineers, 2018-01-02) Javidnia, Hossein; Bazrafkan, Shabab; Corcoran, Peter; Science Foundation Ireland; FotoNation LtdConsumer-level multi-array cameras are a key enabling technology for next generation smartphones imaging systems. The present paper aims to analyze the accuracy of the depth estimation while using different camera combinations in a multi-array camera. This is done by providing a framework of deep neural networks to determine depth map from a sequence of images captured by a multi-array camera. Capturing depth information enables users to perform a range of post-capture edits such as refocusing, and creating a 3D model of any scene. Thus it is essential to calculate an accurate depth map while using the minimum computational resources.