YONA FALINIE A. GAUS

I am a post-doctoral research associate at Department of Computer Science, Durham University working with Prof. Toby Breckon. I work at the intersection of image processing, computer vision, image understanding and machine learning. My research revolves around topics such as object detection, object recognition, object tracking, anomaly detection, in multiple modalities (visible, thermal, xray, drones) imagery.

I publish my research in the computer vision conference where I am also an active reviewer. Recently, I co-organized workshops on anomaly detection (VAND@CVPR2024) in CVPR 2024.

E-Mail Google Scholar Twitter LinkedIn
Towards Open-World Object-based Anomaly Detection via Self-Supervised Outlier SynthesisECCV 2024

Brian K. S. Isaac-Medina, Yona Falinie A. Gaus, Neelanjan Bhowmik,Toby P. Breckon

TL;DR: detecting unseen instances by leveraging an open-world object detector and an out-of-distribution detector via virtual outlier synthesis

arXiv pdf   code project page
Performance Evaluation of Segment Anything Model with Variational Prompting for Application to Non-Visible Spectrum ImageryCVPR Workshop 2024

Yona Falinie A. Gaus, Neelanjan Bhowmik, Brian K. S. Isaac-Medina, Toby P. Breckon

TL;DR: access the effectiveness of SAM via three different prompting strategies applied on X-ray security and thermal imagery

arXiv pdf   code project page
Unaligned 2D to 3D Translation with Conditional Vector-Quantized Code Diffusion using Transformers ICCV 2023

Abril Corona-Figueroa, Sam Bond-Taylor, Neelanjan Bhowmik, Yona Falinie A. Gaus, Toby P. Breckon, Hubert P. H. Shum, Chris G. Willcocks

TL;DR: a simple and novel 2D to 3D synthesis via transformers, where discrete compressed space allows for both fast and high-resolution image synthesis and can be trained independently without requiring aligned datasets

arXiv pdf poster video project page
Region-based Appearance and Flow Characteristics for Anomaly Detection in Infrared Surveillance Imagery CVPR Workshop 2023

Yona Falinie A. Gaus, Neelanjan Bhowmik, Brian K. S. Isaac-Medina, Hubert P. H. Shum, Amir Atapour-Abarghouei, Toby P. Breckon

TL;DR: leveraging both visual appearance and localized motion characteristics, derived from optic flow for anomaly detection under thermal imagery

arXiv pdf video poster
Robust Semi-Supervised Anomaly Detection via Adversarially Learned Continuous Noise CorruptionVISAPP 2023

Jack W. Barker, Neelanjan Bhowmik, Yona Falinie A. Gaus, Toby P. Breckon

TL;DR: simple approach where DAE is optimised to reconstruct the denoised input from the noised input,vastly improve performance on 'leave-one-out' and challenging anomaly detection tasks

arXiv pdf
Lost in Compression: The Impact of Lossy Image Compression on Variable Size Object Detection Within Infrared ImageryCVPR Workshop 2022

Neelanjan Bhowmik, Jack W. Barker, Yona Falinie A. Gaus, Toby P. Breckon

TL;DR: re-training models on lossy JPEG compression notably ameliorated performances, tiny and small objects are more sensitive to compression than medium and large object

arXiv pdf video
Operationalizing Convolutional Neural Network Architectures for Prohibited Object Detection in X-Ray ImageryICMLA 2021

Thomas W. Webb, Neelanjan Bhowmik, Yona Falinie A. Gaus, Toby P. Breckon

TL;DR: retraining the networks on lossy compression reduce storage overhead for deployment within X-ray scanner applications

arXiv pdf video
Continuous Multi-modal Emotion Prediction in Video based on Recurrent Neural Network Variants with Attention ICMLA 2021

Joyal Raju, Yona Falinie A. Gaus, Toby P. Breckon

TL;DR: investigate the SOTA deep learning and its variants to bridge the gap across different culture in emotion prediction

arXiv pdf video
On the impact of using X-Ray Energy Response Imagery for Object Detection via Convolutional Neural NetworksICIP 2021

Neelanjan Bhowmik, Yona Falinie A. Gaus, Toby P. Breckon

TL;DR: training using combination of RGB with energy response and effective-z gives maximal performance over 6-class prohibited item detection problem

arXiv pdf poster video
Visible to Infrared Transfer Learning as a Paradigm for Accessible Real-time Object Detection and Classification in Infrared ImagerySPIE 2020

Yona Falinie A. Gaus, Neelanjan Bhowmik, Brian K.S. Isaac-Medina, and Toby P. Breckon

TL;DR: leveraging prior training on visible-band image datasets, and then subsequently only requiring a secondary, smaller volume of infrared-band imagery for CNN model fine-tuning

arXiv pdf video
On the Impact of Object and Sub-component Level Segmentation Strategies for Supervised Anomaly Detection within X-ray Security Imagery ICMLA 2019

Neelanjan Bhowmik, Yona Falinie A. Gaus, Samet Akcay, Jack W. Barker, Toby P. Breckon

TL;DR: demonstrates the superiority of a varying segmentation (object and sub-component) plus fine-grain CNN architectures

arXiv pdf video
Evaluating the Transferability and Adversarial Discrimination of Convolutional Neural Networks for Threat Object Detection and Classification within X-Ray Security Imagery ICMLA 2019

Yona Falinie A. Gaus, Neelanjan Bhowmik, Samet Akcay, Toby P. Breckon

TL;DR: exploration of CNN transferability between X-ray scanner with differing geometric, resolution and colour profile, generate adversarial dataset for firearms detection

arXiv pdf poster video
On the Use of Deep Learning for the Detection of Firearms in X-ray Baggage Security Imagery HST 2019

Yona Falinie A. Gaus, Neelanjan Bhowmik, Toby P. Breckon

TL;DR: evaluate CNN object detection over two-cass prohibited item detection in X-ray imagery, one-stage detector offer low FP while two-stage achieves high TP

arXiv pdf poster
Using Deep Neural Networks to Address the Evolving Challenges of Concealed Threat Detection within Complex Electronic Items HST 2019

Neelanjan Bhowmik, Yona Falinie A. Gaus, Toby P. Breckon

TL;DR: offers the automatic first-stage screening of aviation baggage for anomalous electronic item detection at the component level as an indicator of potential threat presence

arXiv pdf
The Good, the Bad and the Ugly: Evaluating Convolutional Neural Networks for Prohibited Item Detection Using Real and Synthetically Composited X-ray Imagery BMVC Workshop 2019

Neelanjan Bhowmik, Qian Wang, Yona Falinie A. Gaus, Marcin Szarek, Toby P. Breckon

TL;DR: generate high-quality synthetically composited X-ray images using TIP approach, extensive comparison on how real and synthetic X-ray affects the performance of CNN architecture

arXiv
Evaluation of a Dual Convolutional Neural Network Architecture for Object-wise Anomaly Detection in Cluttered X-ray Security ImageryICMLA 2019

Yona Falinie A. Gaus, Neelanjan Bhowmik, Samet Akcay, Paolo M. Guillen-Garcia, Jack W. Barker, Toby P. Breckon

TL;DR: combine CNN object and fine-grained approaches for object anomaly detection problem, FP remains a significant issue

arXiv pdf
Artificial intelligence system for continuous affect estimation from naturalistic human expressions PhD thesis

Yona Falinie A. Gaus

TL;DR: summary of my work prior to 2018, learning continuous emotion recognition using feature representation and machine learning strategy.

thesis
Linear and Non-Linear Multimodal Fusion for Continuous Affect Estimation In-the-Wild FG 2018

Yona Falinie A Gaus, Hongying Meng

TL;DR: combine audio, video and text by quantify the relationship within these modality via multimodal fusion approaches,

arXiv pdf
Holoscopic 3D Micro-Gesture Database for Wearable Device Interaction FG 2018

Yi Liu, Hongying Meng, Mohammad Rafiq Swash,Yona Falinie A. Gaus,Rui Qin

TL;DR: unique holoscopic 3D micro-gesture database (HoMG) which supports robust 3D depth micro lens array is created.

arXiv dataset pdf
Artificial Intelligent System for Automatic Depression Level Analysis Through Visual and Vocal ExpressionsIEEE Transaction 2018

Asim Jan, Hongying Meng, Yona Falinie A. Gaus, Fan Zhang

TL;DR: feature extraction via deep learning on audio and video then fused together to capture temporal movement on feature space.

pdf
Decoupling Temporal Dynamics for Naturalistic Affect Recognition in a Two-Stage Regression Framework CYBConf 2017

Yona Falinie A. Gaus, Hongying Meng, Asim Jan

TL;DR: 2-stage regression approach can match or outperform SVR models on naturalistic affect recognition task

arXiv pdf
Social Touch Gesture Recognition using Random Forest and Boosting on Distinct Feature SetsICMI 2015

Yona Falinie A. Gaus, Temitayo Olugbade, Asim Jan, Rui Qin, Jingxin Liu, Fan Zhang, Hongying Meng, Nadia Bianchi-Berthouze

TL;DR: classification between different touch behaviour via ensemble of learning methods such as Random Forest and Boosting

arXiv pdf
Automatic Affective Dimension Recognition from Naturalistic Facial Expressions Based on Wavelet Filtering and PLS RegressionFG Workshop 2015

Yona Falinie A. Gaus, Hongying Meng, Asim Jan, Fan Zhang, and Saeed Turabzadeh

TL;DR: Haar wavelet transform was used to remove noise, fusion process was used for combining video and audio modalities

paper pdf video
Automatic Depression Scale Prediction using Facial Expression Dynamics and Regression ACM Multimedia 2014

Asim Jan, Hongying Meng, Yona Falinie A. Gaus, Fan Zhang, Saeed Turabzadeh

TL;DR: 1-D MHH applied on dynamic feature for the video, PLS and Linear regression were used to capture correlation between feature space and depression scales

paper pdf

Feel free to use this website as template. Inspired by Jon Barron's iconic template.