Browsing by Author "Wang, Tsaipei"

Now showing 1 - 19 of 19

A Flexible Possibilistic C-Template Shell Clustering Method with Adjustable Degree of Deformation
(2016)
Wang, Tsaipei
;
資訊工程學系
;
Department of Computer Science
This paper presents a new method of template based shell clustering that allows more flexible free deformation of the cluster prototypes with respect to the template-defined shapes. This is achieved via a soft division of the template into several template parts, each allowed to have its own set of transform parameters. A fuzzification factor, inspired by the one used in the standard fuzzy c-means algorithm, is used to control the degree of deformation by blending the transform parameters of the template parts. We demonstrate that this approach gives better shape detection and fitting results than the original possibilistic c-template algorithm using synthetic data of several different shapes.
A generalized possibilistic approach to shell clustering of template-based shapes
(2017-02)
Wang, Tsaipei
;
Hung, Wen-Liang
;
資訊工程學系
;
Department of Computer Science
Efficiently and effectively detecting shell-like structures of particular shapes is an important task in computer vision and image processing. This paper presents a generalized possibilistic c-means algorithm (PCM) for shell clustering based on the diversity index of degree-lambda proposed by Patil and Taillie [Diversity as a concept and its measurement. J Amer Statist Assoc. 1982;77:548-561]. Experiments on various data sets in Wang [Possibilistic shell clustering of template-based shapes. IEEE Trans Fuzzy Syst. 2009;17:777-793] show that the the proposed generalized PCM performs better than Wang\'s [Possibilistic shell clustering of template-based shapes. IEEE Trans Fuzzy Syst. 2009;17:777-793] possibilistic shell clustering method according two two criteria: (i) the \'grade of detection\' g(d) for each target cluster; (ii) the amount of computation, denoted as k(c), required to attain a given g(d).
Automatic segmentation and summarization for videos taken with smart glasses
(2018-05-01)
Chiu, Yen-Chia
;
Liu, Li-Yi
;
Wang, Tsaipei
;
資訊工程學系
;
Department of Computer Science
This paper discusses the topic of automatic segmentation and extraction of important segments of videos taken with Google Glasses. Using the information from both the video images and additional sensor data that are recorded concurrently, we devise methods that automatically divide the video into coherent segments and estimate the importance of the each segment. Such information then enables automatic generation of video summary that contains only the important segments. The features used include colors, image details, motions, and speeches. We then train multi-layer perceptrons for the two tasks (segmentation and importance estimation) according to human annotations. We also present a systematic evaluation procedure that compares the automatic segmentation and importance estimation results with those given by multiple users and demonstrate the effectiveness of our approach.
CA-Tree: A Hierarchical Structure for Efficient and Scalable Coassociation-Based Cluster Ensembles
(2011-06-01)
Wang, Tsaipei
;
資訊工程學系
;
Department of Computer Science
Chromosome classification based on the band profile similarity along approximate medial axis
(2008-01-01)
Kao, Jau-hong
;
Chuang, Jen-hui
;
Wang, Tsaipei
;
資訊工程學系
;
Department of Computer Science
Automated chromosome classification is an essential task in cytogenetics and has been an important pattern recognition problem. Numerous attempts were made in the past to characterize chromosomes for the purposes of clinical and cancer cytogenetics research. It is important to determine good features and develop feature extraction schemes for chromosome classification. In this paper we propose efficient approaches for medial axis determination and profile matching of human chromosomes without identifying centromeres. The medial axis determination is based on simple cross-section analysis. The features of the band profile obtained along the axis are then used to classify a chromosome based on a subsequence matching technique. Using a special indexing structure, we are able to perform fast similarity search and dynamic insertion and deletion over the established subsequence database of chromosome profiles. According to the experimental results, the developed adaptive system can automatically and efficiently determine the medial axis of a given chromosomes, and achieve satisfactory classification results. (C) 2007 Pattern Recognition Society. Published by Elsevier Ltd. All rights reserved.
Comparing Hard and Fuzzy C-Means for Evidence-Accumulation Clustering
(2009)
Wang, Tsaipei
;
資訊工程學系
;
Department of Computer Science
There exist a multitude of fuzzy clustering algorithms with well understood properties and benefits in various applications. However, there has been very little analysis on using fuzzy clustering algorithms to generate the base clusterings in cluster ensembles. This paper focuses on the comparison of using hard and fuzzy c-means algorithms in the well known evidence-accumulation framework of cluster ensembles. Our new findings include the observations that the fuzzy c-means requires much fewer base clusterings for the cluster ensemble to converge, and is more tolerant of outliers in the data. Some insights are provided regarding the observed phenomena in our experiments.
Driving Style Imitation in Simulated Car Racing Using Style Evaluators and Multi-objective Evolution of a Fuzzy Logic Controller
(2014-01-01)
Wang, Tsaipei
;
Liaw, Keng-Te
;
資訊工程學系
;
Department of Computer Science
This paper describes a new approach to driving style imitation in simulated car racing games. Our goal is to be able to create non-personal characters (NPCs) that both run competitively and exhibit some driving style traits of the player being imitated. We introduce a style evaluator function that can measure the style similarity between driving records even from different tracks. The effectiveness of such style evaluators are verified using driving records of both NPCs and human players. To build NPC drivers that can imitate particular human players, we use a base driver AI based on a fuzzy logic controller and optimizes its parameters using multi-objective evolution. This is the first work on driver imitation that actually allows several human players to drive in their only natural, not instructed, styles. Our results show evidences that the created imitator NPCs do possess traits of styles of the respective human players being modeled.
Frequency subband processing and feature analysis of forward-looking ground-penetrating radar signals for land-mine detection
(2007-03-01)
Wang, Tsaipei
;
Keller, James M.
;
Gader, Paul D.
;
Sjahputera, Ozy
;
資訊工程學系
There has been significant amount of study on the use of ground-penetrating radar (GPR) for land-mine detection. This paper presents our analysis of GPR data collected at a U.S. Army test site using a new approach based on frequency subband processing. In this approach, from the radar data that have over 2.5 GHz of bandwidth, we compute separate radar images using the one wide (2 GHz) and four narrow (0.6 GHz) frequency subbands. The results indicate that signals for different frequency subbands; are significantly different and give very different performance in land-mine detection. In addition, we also examine a number of features extracted from the GPR data, including magnitude and local-contrast features, ratio between copolarization and cross-polarization signals, and features obtained using polarimetric decomposition. Feature selection procedures are employed to find subsets of features that improve detection performance when combined. Results of land-mine detection, including performance on blind test lanes, are presented.
Identifying and Filling Occlusion Holes on Planar Surfaces for 3-D Scene Editing
(2017-01-01)
Wang, Chia-Wei
;
Yang, Tzu-Chieh
;
Chiang, Sheng-Ho
;
Wang, Tsaipei
;
交大名義發表
While it is very convenient to acquire 3-D models of indoor environments today, to edit such an environment is not as easy, as the moving of any object will likely result in exposed occlusion holes. This paper presents a method of automatically identifying and filling such occlusion holes on planar surfaces in 3-D scenes. Steps of hole extraction, classification, and filling using a 2-D image inpainting technique are discussed.
ITERATIVE DATA-DRIVEN CORONARY VESSEL LABELING
(2017-01-01)
Wang, Tsaipei
;
資訊工程學系
;
Department of Computer Science
This paper describes an iterative data-driven algorithm for automatically labeling coronary vessel segments in MDCT images. Such techniques are useful for effective presentation and communication of findings on coronary vessel pathology by physicians and computer-assisted diagnosis systems. The experiments are done on the 18 sets of coronary vessel data in the Rotterdam Coronary Artery Algorithm Evaluation Framework that contain segment labeling by medical experts. The performance of our algorithm show both good accuracy and efficiency compared to previous works on this task.
Learning Overtaking and Blocking Skills in Simulated Car Racing
(2015)
Huang, Han-Hsien
;
Wang, Tsaipei
;
資訊工程學系
;
Department of Computer Science
In this paper we describe the analysis of using Q-learning to acquire overtaking and blocking skills in simulated car racing games. Overtaking and blocking are more complicated racing skills compared to driving alone, and past work on this topic has only touched overtaking in very limited scenarios. Our work demonstrates that a driving AI agent can learn overtaking and blocking skills via machine learning, and the acquired skills are applicable when facing different opponent types and track characteristics, even on actual built-in tracks in TORCS.
Mask-RCNN Based People Detection Using A Top-View Fisheye Camera
(2019-01-01)
Wang, Tsaipei
;
Hsieh, Yun-Yi
;
Wong, Fong-Wen
;
Chen, Yi-Fu
;
資訊工程學系
People detection is a core problem in many applications, such as those related to visual surveillance and behavior analysis. In recent years, we have seen dramatic improvements of people detection techniques based on deep neural networks. However, most of the techniques are developed for projective cameras where people mostly appear upright in the images. On the other hand, images taken with top-view fisheye cameras, another major modality used in visual surveillance, have received little attention. In this paper, we present a study on extending the well-known Mask-RCNN people detection algorithm to such images.
ON THE EFFECT OF DATA IMBALANCE FOR MULTI-LABEL PEDESTRIAN ATTRIBUTE RECOGNITION
(2018-01-01)
Wang, Tsaipei
;
Shu, Kai-Chen
;
Chang, Chia-Hao
;
Chen, Yi-Fu
;
資訊工程學系
Pedestrian attribute recognition has many applications in surveillance and attribute based query, tracking, and person re-identification. The recent trend in deep-learning based pedestrian attribute recognition is to use a shared CNN backbone for feature extraction and multiple subsequent branches for the individual branches. While this allows the end-to-end learning to simultaneously recognize multiple attributes, the data imbalance problem of most attributes becomes a challenge that has not been studied sufficiently for this application. This paper presents studies on how the cost adjustment method affects several common evaluation metrics. We also propose a two-stage training procedure, where an additional fine-tuning stage on the classifier layers only with class-balanced data is shown to improve recognition performances.
Possibilistic c-template clustering and its application in object detection in images
(2006)
Wang, Tsaipei
;
資訊工程學系
;
Department of Computer Science
We present in this paper a new type of alternating-optimization based possibilistic c-shell clustering algorithm called possibilistic c-template (PCT). A template is represented by a set of line segments. A cluster prototype consists of a copy of the template after translation, scaling, and rotation transforms. This extends the capability of shell clustering beyond a few standard geometrical shapes that have been studied so far. We use a number of 2-dimensional data sets to illustrate the application of our algorithm in detecting generic template-based shapes in images. Techniques taken to relax the requirements of known number of clusters and good initialization are also described. Results for both synthetic and actual image data are presented.
Possibilistic Clustering of Generic Shapes Derived from Templates
(2008)
Wang, Tsaipei
;
資訊工程學系
;
Department of Computer Science
We present in this paper a new type of alternating-optimization based possibilistic c-shell algorithm for clustering template-based shapes. A cluster prototype consists of a copy of the template after translation, seating, rotation, and/or affine transformations. We use a number of two-dimensional data sets, both synthetic and from real-world images, to illustrate the capability of our algorithm in detecting generic template-based shapes in images. We also describe a progressive clustering procedure aimed to relax the requirements of known number of dusters and good initialization.
Possibilistic Shell Clustering of Template-Based Shapes
(2009-08-01)
Wang, Tsaipei
;
資訊工程學系
;
Department of Computer Science
In this paper, we present a new type of alternating-optimization- based possibilistic c-shell algorithm for clustering-template- based shapes. A cluster prototype consists of a copy of the template after translation, scaling, rotation, and/or affine transformations. This extends the capability of shell clustering beyond a few standard geometrical shapes that have been in the literature so far. We use a number of 2-D datasets, consisting of both synthetic and real-world images, to illustrate the capability of our algorithm in detecting generic-template-based shapes in images. We also describe a progressive clustering procedure aimed to relax the requirements for a known number of clusters and good initialization, as well as new performance measures of shell-clustering algorithms.
TEMPLATE-BASED PEOPLE DETECTION USING A SINGLE DOWNWARD-VIEWING FISHEYE CAMERA
(2017-01-01)
Wang, Tsaipei
;
Chang, Chia-Wei
;
Wu, Yu-Shan
;
資訊工程學系
;
Department of Computer Science
This paper describes a new algorithm for detecting people using a single downward-viewing fisheye camera. People detection from images taken with projective cameras has been studied extensively in recent years. On the other hand, researches on people detection from fisheye camera images are very limited, and existing techniques are either designed for very simplistic and uncluttered environments, or are very time-consuming. Given that fisheye cameras have a number of advantages in application such as people counting and visual surveillance, including less occlusion among peoples and larger views, the objective here is to propose a new technique of people counting using fisheye cameras that is both practical to realistic environments and efficient enough for real-time applications. The main innovation of our method is to take advantage of fisheye geometries to estimate the expected sizes of people as functions of image locations. An adaptive set of elliptic templates are pre computed to expedite processing. Given the different appearances of people at different distances to the image center, a set of support vector machines (SVMs) are used to classify different templates as people or not. We also describe a tracking algorithm for people counting and tracking in indoor environments.
Template-Based Shell Clustering Using a Line-Segment Representation of Data
(2011-06-01)
Wang, Tsaipei
;
資訊工程學系
;
Department of Computer Science
This paper presents the algorithms and experimental results for template-based shell clustering when the datasets are represented by line segments. Compared with point datasets, such representations have several advantages, which include better scalability and noise immunity, as well as the availability of orientation information. Using both synthetic and real-world image datasets, we have experimentally demonstrated that line-segment-based representations result in both better accuracy and better efficiency in shell clustering.
The Development of a Simulated Car Racing Controller Based on Monte-Carlo Tree Search
(2016-01-01)
Hou, Jia-Hao
;
Wang, Tsaipei
;
資訊工程學系
;
Department of Computer Science
Ever since its introduction, Monte Carlo Tree Search (MCTS) has shown very good performances on a number of games, most of which are turn-based zero-sum games. More recently, researchers have also started to expand the application of MCTS to other types of games. This paper proposes a new framework of applying MCTS to the game of simulated car racing. We choose to build the search tree in a discretized game-state space and then determine the action from the selected target game state. This allows us to avoid the need to discretize the action space. In addition, we are able to incorporate some heuristics on driving strategies naturally. The resulting controller shows very competitive performance in the open-source racing game TORCS.