Engineering Commons^™

Evaluating The Performance Of Vision Transformer Architecture For Deepfake Image Classification, Devesan Govindasamy

Dissertations

Deepfake classification has seen some impressive results lately, with the experimentation of various deep learning methodologies, researchers were able to design some state-of-the art techniques. This study attempts to use an existing technology “Transformers” in the field of Natural Language Processing (NLP) which has been a de-facto standard in text processing for the purposes of Computer Vision. Transformers use a mechanism called “self-attention”, which is different from CNN and LSTM. This study uses a novel technique that considers images as 16x16 words (Dosovitskiy et al., 2021) to train a deep neural network with “self-attention” blocks to detect deepfakes. It creates …

Gradient Free Sign Activation Zero One Loss Neural Networks For Adversarially Robust Classification, Yunzhe Xue

Dissertations

The zero-one loss function is less sensitive to outliers than convex surrogate losses such as hinge and cross-entropy. However, as a non-convex function, it has a large number of local minima, andits undifferentiable attribute makes it impossible to use backpropagation, a method widely used in training current state-of-the-art neural networks. When zero-one loss is applied to deep neural networks, the entire training process becomes challenging. On the other hand, a massive non-unique solution probably also brings different decision boundaries when optimizing zero-one loss, making it possible to fight against transferable adversarial examples, which is a common weakness in deep learning …

Towards Adversarial Robustness With 01 Lossmodels, And Novel Convolutional Neural Netsystems For Ultrasound Images, Meiyan Xie

Dissertations

This dissertation investigates adversarial robustness with 01 loss models and a novel convolutional neural net systems for vascular ultrasound images.

In the first part, the dissertation presents stochastic coordinate descent for 01 loss and its sensitivity to adversarial attacks. The study here suggests that 01 loss may be more resilient to adversarial attacks than the hinge loss and further work is required.

In the second part, this dissertation proposes sign activation network with a novel gradient-free stochastic coordinate descent algorithm and its ensembling model. The study here finds that the ensembling model gives a high minimum distortion (as measured by …

Development Of Deep Learning Neural Network For Ecological And Medical Images, Shaobo Liu

Dissertations

Deep learning in computer vision and image processing has attracted attentions from various fields including ecology and medical image. Ecologists are interested in finding an effective model structure to classify different species. Tradition deep learning model use a convolutional neural network, such as LeNet, AlexNet, VGG models, residual neural network, and inception models, are first used on classifying bee wing and butterfly datasets. However, insufficient data sample and unbalanced samples in each class have caused a poor accuracy. To make improvement the test accuracy, data augmentation and transfer learning are applied. Recently developed deep learning framework based on mathematical morphology …

Deep Learning On Image Forensics And Anti-Forensics, Zhangyi Shen

Dissertations

Image forensics protect the authenticity and integrity of digital images. On the contrary, as the countermeasures of digital forensics, anti-forensics is applied to expose the vulnerability of forensics tools. Consequently, forensics researchers could develop forensics tools against possible new attacks. This dissertation investigation demonstrates two image forensics methods based on convolutional neural network (CNN) and two image anti-forensics methods based on generative adversarial network (GAN).

Detecting unsharp masking (USM) sharpened image is the first study in this dissertation. A CNN architecture comprises four convolutional layers and a classification module is proposed to discriminate sharpened images and unsharpened images. The results …

Evaluating The Performance Of Transformer Architecture Over Attention Architecture On Image Captioning, Deepti Balasubramaniam

Dissertations

Over the last few decades computer vision and Natural Language processing has shown tremendous improvement in different tasks such as image captioning, video captioning, machine translation etc using deep learning models. However, there were not much researches related to image captioning based on transformers and how it outperforms other models that were implemented for image captioning. In this study will be designing a simple encoder-decoder model, attention model and transformer model for image captioning using Flickr8K dataset where will be discussing about the hyperparameters of the model, type of pre-trained model used and how long the model has been trained. …

A Deep Machine Learning Approach For Predicting Freeway Work Zone Delay Using Big Data, Abdullah Shabarek

Dissertations

The introduction of deep learning and big data analytics may significantly elevate the performance of traffic speed prediction. Work zones become one of the most critical factors causing congestion impact, which reduces the mobility as well as traffic safety. A comprehensive literature review on existing work zone delay prediction models (i.e., parametric, simulation and non-parametric models) is conducted in this research. The research shows the limitations of each model. Moreover, most previous modeling approaches did not consider user delay for connected freeways when predicting traffic speed under work zone conditions. This research proposes Deep Artificial Neural Network (Deep ANN) and …

Transformer Neural Networks For Automated Story Generation, Kemal Araz

Dissertations

Towards the last two-decade Artificial Intelligence (AI) proved its use on tasks such as image recognition, natural language processing, automated driving. As discussed in the Moore’s law the computational power increased rapidly over the few decades (Moore, 1965) and made it possible to use the techniques which were computationally expensive. These techniques include Deep Learning (DL) changed the field of AI and outperformed other models in a lot of fields some of which mentioned above. However, in natural language generation especially for creative tasks that needs the artificial intelligent models to have not only a precise understanding of the given …

Multi-Sensory Deep Learning Architectures For Slam Dunk Scene Classification, Paul Minogue

Dissertations

Basketball teams at all levels of the game invest a considerable amount of time and effort into collecting, segmenting, and analysing footage from their upcoming opponents previous games. This analysis helps teams identify and exploit the potential weaknesses of their opponents and is commonly cited as one of the key elements required to achieve success in the modern game. The growing importance of this type of analysis has prompted research into the application of computer vision and audio classification techniques to help teams classify scoring sequences and key events using game footage. However, this research tends to focus on classifying …

Computational Intelligence In Steganography: Adaptive Image Watermarking, Xin Zhong

Dissertations

Digital image watermarking, as an extension of traditional steganography, refers to the process of hiding certain messages into cover images. The transport image, called marked-image or stego-image, conveys the hidden messages while appears visibly similar to the cover-image. Therefore, image watermarking enables various applications such as copyright protection and covert communication. In a watermarking scheme, fidelity, capacity and robustness are considered as crucial factors, where fidelity measures the similarity between the cover- and marked-images, capacity measures the maximum amount of watermark that can be embedded, and robustness concerns the watermark extraction under attacks on the marked-image. Watermarking techniques are often …