Digital Commons Network^™

Generalized Differentiable Neural Architecture Search With Performance And Stability Improvements, Emily J. Herron

Doctoral Dissertations

This work introduces improvements to the stability and generalizability of Cyclic DARTS (CDARTS). CDARTS is a Differentiable Architecture Search (DARTS)-based approach to neural architecture search (NAS) that uses a cyclic feedback mechanism to train search and evaluation networks concurrently, thereby optimizing the search process by enforcing that the networks produce similar outputs. However, the dissimilarity between the loss functions used by the evaluation networks during the search and retraining phases results in a search-phase evaluation network, a sub-optimal proxy for the final evaluation network utilized during retraining. ICDARTS, a revised algorithm that reformulates the search phase loss functions to ensure …

Attention Mechanism In Deep Neural Networks For Computer Vision Tasks, Haohan Li

Doctoral Dissertations

“Attention mechanism, which is one of the most important algorithms in the deep Learning community, was initially designed in the natural language processing for enhancing the feature representation of key sentence fragments over the context. In recent years, the attention mechanism has been widely adopted in solving computer vision tasks by guiding deep neural networks (DNNs) to focus on specific image features for better understanding the semantic information of the image. However, the attention mechanism is not only capable of helping DNNs understand semantics, but also useful for the feature fusion, visual cue discovering, and temporal information selection, which are …

Detecting Cells And Analyzing Their Behaviors In Microscopy Images Using Deep Neural Networks, Yunxiang Mao

Doctoral Dissertations

"The computer-aided analysis in the medical imaging field has attracted a lot of attention for the past decade. The goal of computer-vision based medical image analysis is to provide automated tools to relieve the burden of human experts such as radiologists and physicians. More specifically, these computer-aided methods are to help identify, classify and quantify patterns in medical images. Recent advances in machine learning, more specifically, in the way of deep learning, have made a big leap to boost the performance of various medical applications. The fundamental core of these advances is exploiting hierarchical feature representations by various deep learning …

Detecting, Segmenting And Tracking Bio-Medical Objects, Mingzhong Li

Doctoral Dissertations

"Studying the behavior patterns of biomedical objects helps scientists understand the underlying mechanisms. With computer vision techniques, automated monitoring can be implemented for efficient and effective analysis in biomedical studies. Promising applications have been carried out in various research topics, including insect group monitoring, malignant cell detection and segmentation, human organ segmentation and nano-particle tracking.

In general, applications of computer vision techniques in monitoring biomedical objects include the following stages: detection, segmentation and tracking. Challenges in each stage will potentially lead to unsatisfactory results of automated monitoring. These challenges include different foreground-background contrast, fast motion blur, clutter, object overlap and …