Current Search: Liu, Fei (x)
-
-
Title
-
Human Action Detection, Tracking and Segmentation in Videos.
-
Creator
-
Tian, Yicong, Shah, Mubarak, Bagci, Ulas, Liu, Fei, Walker, John, University of Central Florida
-
Abstract / Description
-
This dissertation addresses the problem of human action detection, human tracking and segmentation in videos. They are fundamental tasks in computer vision and are extremely challenging to solve in realistic videos. We first propose a novel approach for action detection by exploring the generalization of deformable part models from 2D images to 3D spatiotemporal volumes. By focusing on the most distinctive parts of each action, our models adapt to intra-class variation and show robustness to...
Show moreThis dissertation addresses the problem of human action detection, human tracking and segmentation in videos. They are fundamental tasks in computer vision and are extremely challenging to solve in realistic videos. We first propose a novel approach for action detection by exploring the generalization of deformable part models from 2D images to 3D spatiotemporal volumes. By focusing on the most distinctive parts of each action, our models adapt to intra-class variation and show robustness to clutter. This approach deals with detecting action performed by a single person. When there are multiple humans in the scene, humans need to be segmented and tracked from frame to frame before action recognition can be performed. Next, we propose a novel approach for multiple object tracking (MOT) by formulating detection and data association in one framework. Our method allows us to overcome the confinements of data association based MOT approaches, where the performance is dependent on the object detection results provided at input level. We show that automatically detecting and tracking targets in a single framework can help resolve the ambiguities due to frequent occlusion and heavy articulation of targets. In this tracker, targets are represented by bounding boxes, which is a coarse representation. However, pixel-wise object segmentation provides fine level information, which is desirable for later tasks. Finally, we propose a tracker that simultaneously solves three main problems: detection, data association and segmentation. This is especially important because the output of each of those three problems are highly correlated and the solution of one can greatly help improve the others. The proposed approach achieves more accurate segmentation results and also helps better resolve typical difficulties in multiple target tracking, such as occlusion, ID-switch and track drifting.
Show less
-
Date Issued
-
2018
-
Identifier
-
CFE0007378, ucf:52069
-
Format
-
Document (PDF)
-
PURL
-
http://purl.flvc.org/ucf/fd/CFE0007378
-
-
Title
-
Learning Internal State Memory Representations from Observation.
-
Creator
-
Wong, Josiah, Gonzalez, Avelino, Liu, Fei, Wu, Annie, Ontanon, Santiago, Wiegand, Rudolf, University of Central Florida
-
Abstract / Description
-
Learning from Observation (LfO) is a machine learning paradigm that mimics how people learn in daily life: learning how to do something simply by watching someone else do it. LfO has been used in various applications, from video game agent creation to driving a car, but it has always been limited by the inability of an observer to know what a performing entity chooses to remember as they act in an environment. Various methods have either ignored the effects of memory or otherwise made...
Show moreLearning from Observation (LfO) is a machine learning paradigm that mimics how people learn in daily life: learning how to do something simply by watching someone else do it. LfO has been used in various applications, from video game agent creation to driving a car, but it has always been limited by the inability of an observer to know what a performing entity chooses to remember as they act in an environment. Various methods have either ignored the effects of memory or otherwise made simplistic assumptions about its structure. In this dissertation, we propose a new method, Memory Composition Learning, that captures the influence of a performer's memory in an observed behavior through the creation of an auxiliary memory feature set that explicitly models the aspects of the environment with significance for future decisions, and which can be used with a machine learning technique to provide salient information from memory. It advances the state of the art by automatically learning the internal structure of memory instead of ignoring or predefining it. This research is difficult in that memory modeling is an unsupervised learning problem that we elect to solve solely from unobtrusive observation. This research is significant for LfO in that it will allow learning techniques that otherwise could not use information from memory to use a tailored set of learned memory features that capture salient influences from memory and enable decision-making based on these influences for more effective learning performance. To validate our hypothesis, we implemented a prototype for modeling observed memory influences with our approach and applied it to simulated vacuum cleaner and lawn mower domains. Our investigation revealed that MCL was able to automatically learn memory features that describe the influences on an observed actor's internal state, and which improved learning performance of observed behaviors.
Show less
-
Date Issued
-
2019
-
Identifier
-
CFE0007879, ucf:52755
-
Format
-
Document (PDF)
-
PURL
-
http://purl.flvc.org/ucf/fd/CFE0007879
-
-
Title
-
Leaning Robust Sequence Features via Dynamic Temporal Pattern Discovery.
-
Creator
-
Hu, Hao, Wang, Liqiang, Zhang, Shaojie, Liu, Fei, Qi, GuoJun, Zhou, Qun, University of Central Florida
-
Abstract / Description
-
As a major type of data, time series possess invaluable latent knowledge for describing the real world and human society. In order to improve the ability of intelligent systems for understanding the world and people, it is critical to design sophisticated machine learning algorithms for extracting robust time series features from such latent knowledge. Motivated by the successful applications of deep learning in computer vision, more and more machine learning researchers put their attentions...
Show moreAs a major type of data, time series possess invaluable latent knowledge for describing the real world and human society. In order to improve the ability of intelligent systems for understanding the world and people, it is critical to design sophisticated machine learning algorithms for extracting robust time series features from such latent knowledge. Motivated by the successful applications of deep learning in computer vision, more and more machine learning researchers put their attentions on the topic of applying deep learning techniques to time series data. However, directly employing current deep models in most time series domains could be problematic. A major reason is that temporal pattern types that current deep models are aiming at are very limited, which cannot meet the requirement of modeling different underlying patterns of data coming from various sources. In this study we address this problem by designing different network structures explicitly based on specific domain knowledge such that we can extract features via most salient temporal patterns. More specifically, we mainly focus on two types of temporal patterns: order patterns and frequency patterns. For order patterns, which are usually related to brain and human activities, we design a hashing-based neural network layer to globally encode the ordinal pattern information into the resultant features. It is further generalized into a specially designed Recurrent Neural Networks (RNN) cell which can learn order patterns in an online fashion. On the other hand, we believe audio-related data such as music and speech can benefit from modeling frequency patterns. Thus, we do so by developing two types of RNN cells. The first type tries to directly learn the long-term dependencies on frequency domain rather than time domain. The second one aims to dynamically filter out the ``noise" frequencies based on temporal contexts. By proposing various deep models based on different domain knowledge and evaluating them on extensive time series tasks, we hope this work can provide inspirations for others and increase the community's interests on the problem of applying deep learning techniques to more time series tasks.
Show less
-
Date Issued
-
2019
-
Identifier
-
CFE0007470, ucf:52679
-
Format
-
Document (PDF)
-
PURL
-
http://purl.flvc.org/ucf/fd/CFE0007470
-
-
Title
-
Learning Kernel-based Approximate Isometries.
-
Creator
-
Sedghi, Mahlagha, Georgiopoulos, Michael, Anagnostopoulos, Georgios, Atia, George, Liu, Fei, University of Central Florida
-
Abstract / Description
-
The increasing availability of public datasets offers an inexperienced opportunity to conduct data-driven studies. Metric Multi-Dimensional Scaling aims to find a low-dimensional embedding of the data, preserving the pairwise dissimilarities amongst the data points in the original space. Along with the visualizability, this dimensionality reduction plays a pivotal role in analyzing and disclosing the hidden structures in the data. This work introduces Sparse Kernel-based Least Squares Multi...
Show moreThe increasing availability of public datasets offers an inexperienced opportunity to conduct data-driven studies. Metric Multi-Dimensional Scaling aims to find a low-dimensional embedding of the data, preserving the pairwise dissimilarities amongst the data points in the original space. Along with the visualizability, this dimensionality reduction plays a pivotal role in analyzing and disclosing the hidden structures in the data. This work introduces Sparse Kernel-based Least Squares Multi-Dimensional Scaling approach for exploratory data analysis and, when desirable, data visualization. We assume our embedding map belongs to a Reproducing Kernel Hilbert Space of vector-valued functions which allows for embeddings of previously unseen data. Also, given appropriate positive-definite kernel functions, it extends the applicability of our methodto non-numerical data. Furthermore, the framework employs Multiple Kernel Learning for implicitlyidentifying an effective feature map and, hence, kernel function. Finally, via the use ofsparsity-promoting regularizers, the technique is capable of embedding data on a, typically, lowerdimensionalmanifold by naturally inferring the embedding dimension from the data itself. In theprocess, key training samples are identified, whose participation in the embedding map's kernelexpansion is most influential. As we will show, such influence may be given interesting interpretations in the context of the data at hand. The resulting multi-kernel learning, non-convex framework can be effectively trained via a block coordinate descent approach, which alternates between an accelerated proximal average method-based iterative majorization for learning the kernel expansion coefficients and a simple quadratic program, which deduces the multiple-kernel learning coefficients. Experimental results showcase potential uses of the proposed framework on artificial data as well as real-world datasets, that underline the merits of our embedding framework. Our method discovers genuine hidden structure in the data, that in case of network data, matches the results of well-known Multi- level Modularity Optimization community structure detection algorithm.
Show less
-
Date Issued
-
2017
-
Identifier
-
CFE0007132, ucf:52315
-
Format
-
Document (PDF)
-
PURL
-
http://purl.flvc.org/ucf/fd/CFE0007132