Classification and Information Extraction for Complex and Nested Tabular Structures in Images

Size: px
Start display at page:

Download "Classification and Information Extraction for Complex and Nested Tabular Structures in Images"

Transcription

1 Classification and Information Extraction for Complex and Nested Tabular Structures in Images Abstract Understanding of technical documents, like manuals, is one of the most important steps in automatic reporting and/or troubleshooting of defects. The majority of the relevant information exists in tabular structure. There are some solutions for extracting tabular structures from text. However, it is still a big issue to extract tabular information from images and, on top of that, from complex and nested tables. This paper aims to propose classification and information extraction methods for complex tabular structures in document images. These are hybrid approaches using both image layout and OCRed text. The proposed methods outperform on a real-world technical documents dataset from a German railway company (Deutsche Bahn AG) as compared to other state-of-the-art approaches. As a result, the proposed approaches won the competition held by Deutsche Bahn AG in 2016 against other participating research groups and companies. Index Terms Classification, Information Extraction, Table Information Extraction, Tabular Structure I. INTRODUCTION Most technical documents, that are used for reporting and/or troubleshooting of problems, exist as documents with tabular layout. For an automation of this process, document understanding is required. Already existing solutions can deal with extracting tables, more specifically tabular structures. Still, performing this process on images that contain complex, nested tables is a problem as most state-of-the-art solutions only work for simple text-based scenarios. In 2016, a German railway company, i.e. Deutsche Bahn AG (DB), held a competition [1] about classifying and extracting information of document images with complex, nested tabular structures. These give rise to many challenges in general, and specifically with respect to both classification and information extraction (IE). The common challenges include: i) Varying image resolutions: From low to high, as can be seen in figures 1(a) and 1(b). ii) Bad optical character recognition (OCR) results even for high resolution images as can be seen in figure 1(c) (which is the OCR result of figure 1(b) using Tesseract [2]). The specific challenges with respect to classification are: i) A high intra-class complexity: Documents that seem to be similar from a point of view in structure and layout, but are belonging to different classes, as can be seen in figure 2(a). ii) Labeled and non-labeled class introductions: As illustrated by figure 2(b), sometimes a class name is introduced by the keyword Benennung, and sometimes not; even within the same class. The specific challenges with respect to IE are: i) A complex and nested layout consisting of multiple tables, like in figures 2(a) to 3. ii) Non-unique labeling: The Contributed equally. same label is used multiple times in one image, as visualized by figure 3. iii) Different reading orders within the same document: As described by figure 3, sometimes a value is above, below, behind, or in front of its corresponding label; tables have to be read bottom-up. iv) Different field sets for the same class, due to intra-class complexity. State-of-the-art approaches for document classification, to be precise general document classification and table type classification, are: i) Image-based approaches [3], [4]. But those will suffer from the high intra-class complexity. ii) Keywords-based approaches [4], which will have problems with labeled and non-labeled class introductions. iii) Text-based approaches [3], [4]. But they will have problems with the different resolutions, and the bad OCR results. iv) Deep learning methods, like a combination of region-based and holistic analysis via a convolutional neural network (CNN) [5]. However, any neural network approach will not work here, due to a too small set of documents and the lack of a proper ground truth (GT). (More details on the latter will be provided later in this paper.) Based on the common and the classification-specific challenges of the above mentioned DB technical document scenario, none of these individual approaches could work well for document classification. For the information extraction, table information extraction approaches as well as general IE methods are suited. Especially approaches for semi-structured documents: Because most fields are introduced by labels, the given documents are not considered unstructured. However, field positions may vary from document to document even within the same class therefore the documents also cannot be seen fully structured. State-of-the-art methods here are: i) regular expression (RegEx)-based approaches [6], [7]: Since many of the needed information is introduced or followed by a label, RegEx-based matching seems promising. Still, the non-unique labeling, the different reading orders, and the different fields sets will cause trouble for any RegEx method. ii) Because the documents consist of tabular structures, region of interest (ROI)-based approaches [8], [9] could help guiding the IE process to the right locations. The different label sets and the non-unique naming will, on the other hand, cause problems for ROI-based approaches. iii) A hybrid ontology-based IE has recently been proposed, as it is showing promising results [10]. Even though it worked for a small amount of sentences, i.e. mostly four sentences, it is not applicable to the here given scenario. The problem is, that in this scenario there are if any only labels; no complete sentences. Same like before, none of these individual state-of-the-art IE approaches are suitable for the

2 (a) Low resolution image (b) High resolution image (c) OCR result of figure 1(b) using Tesseract Fig. 1. Common challenges for document images with complex, nested tabular layout. (a) Similar layout and structure, but different classes. (b) Labeled and non-labeled class introduction Fig. 2. Challenges for the document type classification. Fig. 3. Challenges for the information extraction: Non-unique labeling and different reading orders

3 DB scenario, because of the mentioned challenges. In this paper, hybrid approaches, combining image layout and OCRed text, for both classification and information extraction of document images with complex and nested tabular structures are proposed. These techniques solve the limitations of existing state-of-the-art approaches and also outperform competing approaches in the DB competition from The rest of the paper is organized as follows: Section II is about explaining the here proposed methods in detail. In section III the approaches of other research groups and companies, that also participated in the competition, are presented. Section IV contains a performance evaluation, where all here named methods are evaluated; including the competitor s, the here proposed, and state-of-the-art approaches. This section additionally contains a discussion. The last section, section V, is about summarizing the results and drawing a conclusion. II. THE PROPOSED CLASSIFICATION AND INFORMATION EXTRACTION METHODS FOR COMPLEX TABULAR STRUCTURES IN IMAGES In this section, novel document classification and information extraction techniques, using both document layout and OCRed text in a hybrid manner, are described in detail. A. The Proposed Hybrid Classification Approach The proposed document classification method uses document layout and OCRed text information in an interleaved fashion. The procedure is described here as follows: 1) Increase image contrast to enhance the OCRed text. 2) Extract the biggest rectangles (max. five) using connected components (CC) analysis from the image. The rationale behind this operation is that the class of the document is most probably located within one of these rectangles. If this is not the case, further processing is done as discussed in the next steps. 3) As first trial, get the OCRed text of the rectangles detected using Tesseract. Then find a good match between any string from the generated OCR and any class name given the list of all class names. The matching is done using a threshold, which was experimentally set to 0.6. If more than one good match is found, the one with the highest matching ratio is selected. The process of finding a good match works as follows: First, search for the keyword Benennung, since this is, for some images, followed by the class name. But only looking for this keyword won t be effective enough. If it is found, the string following it is extracted and compared with the list of classes provided. Keeping in mind the given threshold while matching. Otherwise, go through the OCRed text line by line: Whenever a good match is found, store its matching ratio, such that the match with the highest ratio can be chosen at the end. It is important to mention that some class names are prefix of others. Additionally, in some cases the class name is split over two or more different lines. The presented technique takes these observations into account by continuously searching for the longest possible and available class name (a.k.a. greedy matching). 4) As second trial, in case no good match was found, apply image denoising before applying the OCR to produce a better OCRed text. Afterwards, a good match (as defined above) is found between strings from the OCRed text and the given class names; but with a threshold of ) Finally, as last trial, get the OCR of the whole image with increased contrast and also search for a good match (as defined above) with a threshold of 0.5. Otherwise, the image class is assigned to Unknown, which is actually one of the possible classes. B. The Proposed Hybrid information extraction Approach The proposed information extraction method is based on the combination of relative ROIs and RegEx. The steps are described as follows: 1) Detect all non-empty rectangles, using a CC-based approach, in the image with increased contrast and extract them in reading order. The coordinates of the top left corner of each rectangle are saved, in order to get its surrounding rectangles as well as to compare its location with other rectangles. 2) Get the OCRed text for each rectangle using Tesseract. 3) An extraction method is described for each of the desired information fields using a set of RegEx and four functions which can get all rectangles located to the left, right, top, and bottom, with respect to a specific rectangle. 4) Finally, all the fields are extracted using the defined extraction methods over the detected rectangles and their OCRed text. III. COMPETING APPROACHES The DB competition for technical documents classification and information extraction was an open challenge. Including us, ten different companies and research groups from all around Germany participated in the competition. We are thankful to the following two participants, one from the commercial site (Team Hybrid ) and another one from the research site (Team Awesome ), who provided us a brief description of their solutions. In section IV, we are comparing our solution (the proposed methods by this paper) with these participants. However, as a quick summary, our solution got the 1 st position and the teams Hybrid and Awesome got the 3 rd and 5 th positions, respectively. A. Kofax-based Solution The DB Systel company [11] participated in the DB competition under the team name Hybrid. Their system is based on the commercial Kofax [12] Transformation Products. The usual task of these products is to digitize any formal documents like bills, forms, and delivery notes using predefined functionalities including several features like keyword databases, RegEx, classification, and even trainable classifiers. The DB Systel s solution for the DB competition is described here as follows: Kofax image classifier is used to create a

4 layout-based classification of the documents during a preprocessing step. Based on this information, specific classifiers are trained for the ten most represented classes. There is mostly one classifier, which takes the layout relations of the information fields into account. Several scripts are also implemented for extending the most general functionalities of Kofax Transformation with more task-specific ones, such as OCR substitution. Additionally, all documents that couldn t be classified into the ten most represented classes are combined into one abstract class called no match. Then, in order to process the documents that belong to no match class, trainable classifiers are used because of the layout differences in these documents. B. OCR- and Image Processing-based Solution The Computing Services (ZEDAT) [13], a research group of Freie Universität Berlin, also participated in the competition with the team name Awesome. For document classification, a Tesseract-based OCR is applied on the input document images, after adding all the given class names into the dictionary of Tesseract. Then, each text line is compared with the list of given class names using Hamming Distance. The class name corresponding to the best matching string is recorded as the class name for the input document. For the IE, at first the document is split into individual cells using Canny Edges and Hough Lines as separators. Afterwards the Tesseract OCR engine is applied on each cell independently. Then IE is performed for each field in multiple steps. At first, each cell is checked on its own and is searched for a specific keyword of each field and tried to find the corresponding field value. Since this attempt worked for only 10% of the fields, in a second step, fields without any specific keywords are found. Therefore, the structure of the unmatched fields is analyzed and several simple RegEx are defined to find the field values. As a third step, the surrounding cells are checked for supporting keywords. Additionally, the different positions of the field value needed to be taken into account for each field. Combining these three steps, different fields and their values are detected. IV. PERFORMANCE EVALUATION & DISCUSSION This section contains a performance evaluation and its interpretation, for both classification and information extraction. Contained are a comparison in the area of the DB competition, i.e. between the proposed approaches and the competitor s approaches, and a comparison with state-of-the-art approaches. A. The DB Datasets The organizer of the competition, i.e. Deutsche Bahn AG, provided 909 sample documents, a list of 250 different classes, and a set of 55 fields, where each document belongs to a single class and contains a subset of the provided fields. But they did not provide any classification GT for those samples; however, they provided the field values for a subset of 46 samples as a GT for the IE. The data set consisting of 909 samples will from now own be referred to as DB-909-dataset. It is used by the organizers for comparing the results of all participants. Additionally, for comparing the results of the proposed methods with the state-of-the-art approaches (which were not included in the competition), the GT for document classification is manually created for the DB-909-dataset. For the IE the provided 46 samples with their field values are used. This subset of samples will from now own be referred to as DB-46-dataset. B. The DB Competition The DB asked to upload the solution files, i.e. the files containing the classification assignment and the IE results. To avoid problem-agnostic solutions, they computed the accuracies by themselves. The organizer s computed results for classification, information extraction, and a combination of both as a final score, are shown in table I. The here proposed methods scored a 2 nd place in the classification (with a little margin to the 1 st place holder), the 1 st place for the IE, and the 1 st place in the final rank. This shows how promising the proposed solutions are for this area. TABLE I PERFORMANCE COMPARISON (CLASSIFICATION, INFORMATION EXTRACTION, AND FINAL RANK) BETWEEN COMPETITION PARTICIPANTS, USING THE DB-909-dataset, COMPUTED BY THE COMPETITION ORGANIZERS (I.E. THE DB). Solution classification information extraction Accuracy Rank Accuracy Rank final rank proposed % 2 nd * % 1 st ** 1 st DB Systel % 4 th % 4 th 3 rd ZEDAT % 7 th % 3 rd 5 th * 1 st of classification: % ** 2 nd of IE: % C. State-of-the-art comparison In case of the classification task, the proposed method got compared to an image-based classifier using scale-invariant feature transform (SIFT), a keywords-based approach using a bag-of-words, a text-based approach using Logistic Regression and binary feature weights, and a voting-based approach. The latter was using the results of the three previously mentioned methods and combined them through a majority vote. In case of a conflict, the output of the text-based method was used. This decision is based on the fact, that the text-based classifier achieved the best accuracy among those approaches. The comparison was run on the DB-909-dataset in all five cases. As can be seen in table II, the proposed classification approach achieves an accuracy of %. Hence it outperforms its classical competitors by more than doubling their performance. In case of the IE, the proposed approach got compared to a RegEx- and an ROI-based method. All the methods were executed on the DB-46-dataset. The previous run of the classification in order to being able to run the IE using

5 TABLE II CLASSIFICATION PERFORMANCE COMPARISON OF THE PRESENTED METHOD WITH RESPECT TO STATE-OF-THE-ART CLASSIFICATION APPROACHES OVER THE DB-909-dataset, WHICH IS COMPUTED BY THE AUTHORS OF THIS PAPER. Method image- keywords- textvoting proposed based based based approach Accuracy % % % % % the text-based classifier, achieved an accuracy of %. Due to that fairly low accuracy, the IE was executed twice for each the RegEx- and ROI-based approaches: First it was run directly after the classification finished, without adapting the results manually. This is referred to as Auto. in table III. Second, the classification results were manually corrected and then, the IE was run. This is referred to as Manual in the table. As the proposed approach is performing the IE without considering the sample s class, this distinction was not done for this method. The comparison between GT and the respective approach was done using the Levenshtein distance of the overall result set. As table III illustrates, the here proposed method is, with an accuracy of %, the most accurate approach. TABLE III INFORMATION EXTRACTION PERFORMANCE COMPARISON OF THE PRESENTED METHOD WITH RESPECT TO STATE-OF-THE-ART IE APPROACHES OVER THE DB-46-dataset, WHICH IS COMPUTED BY THE AUTHORS OF THIS PAPER. Method RegEx ROI proposed Auto. Manual Auto. Manual approach Accuracy % % % % % V. CONCLUSION This paper presented classification and information extraction (IE) methods for understanding technical document images containing complex and nested tabular structures. Both of the proposed techniques are developed using a hybrid approach: the document classification using both image layout and OCRed text, and the information extraction from nested and complex tabular structure using regions of interest (ROIs) and regular expressions (RegEx). These proposed methods participated in an open challenge organized by Deutsche Bahn AG (DB) for classification and IE from their technical documents dataset of around 900 samples. These samples are from 250 different document classes where each sample contains a subset of 55 different fields. Altogether ten different commercial companies and research groups participated in this competition. According to the results, which were computed by the organizers, our proposed approaches won the competition and outperformed all other nine participating methods. Additionally, after the competition, we also compared our classification and IE approaches with widely used state-ofthe-art classification and IE techniques, respectively, that were not included in the competition. In this case too, our presented methods outperformed all the state-of-the-art approaches. As one can notice in the performance evaluation results of the IE, even though our IE method got the best results as compared to other participants and state-of-the-art approaches, the results are still not good enough. Extracting information from nested and complex tabular structure, like the ones we are handling in this paper, is a very challenging task which cannot be efficiently solved in a fully automated manner. Therefore in near future, our goal is to introduce a semi-automatic IE technique on the basis of our proposed solution which can interactively involve a user in the loop for correction and then learning the behavior of correction for successively automating the correction. To be added after review. ACKNOWLEDGMENT REFERENCES [1] Deutsche Bahn AG. (2016) DB Digital Header:challenge. [Online]. Available: digital-headerchallenge.html [2] tesseract ocr. (2017) Tesseract Open Source OCR Engine. GitHub. [Online]. Available: [3] S. Bukhari and A. Dengel, Visual appearance based document classication methods: Performance evaluation and benchmarking, in th International Conference on Document Analysis and Recognition (ICDAR), Aug 2015, pp [4] N. Chen and D. Blostein, A survey of document image classification: problem statement, classifier architecture and performance evaluation, vol. 10, pp. 1 16, [5] A. W. Harley, A. Ufkes, and K. G. Derpanis, Evaluation of deep convolutional nets for document image classification and retrieval, Feb [6] A. Bartoli, A. D. Lorenzo, E. Medvet, and F. Tarlao, Inference of regular expressions for text extraction from examples, IEEE Transactions on Knowledge and Data Engineering, vol. 28, no. 5, pp , May [7] R. Fagin, B. Kimelfeld, F. Reiss, and S. Vansummeren, A relational framework for information extraction, SIGMOD Rec., vol. 44, no. 4, pp. 5 16, May [Online]. Available: http: //doi.acm.org/ / [8] R. Girshick, Fast r-cnn, in International Conference on Computer Vision (ICCV), [9] T. Gogar, O. Hubacek, and J. Sedivy, Deep neural networks for web page information extraction, in 12th IFIP WG 12.5 International Conference and Workshops, AIAI 2016, Thessaloniki, Greece, September 16-18, 2016, Proceedings. Springer, Cham, 2016, pp [10] F. Gutierrez, D. Dou, S. Fickas, D. Wimalasuriya, and H. Zong, A hybrid ontology-based information extraction system, [11] DB Systel GmbH. [12] Kofax Inc. [13] Die Zentraleinrichtung für Datenverarbeitung (ZEDAT), Freie Universität Berlin.

Deep Tracking: Biologically Inspired Tracking with Deep Convolutional Networks

Deep Tracking: Biologically Inspired Tracking with Deep Convolutional Networks Deep Tracking: Biologically Inspired Tracking with Deep Convolutional Networks Si Chen The George Washington University sichen@gwmail.gwu.edu Meera Hahn Emory University mhahn7@emory.edu Mentor: Afshin

More information

Improved Classification of Known and Unknown Network Traffic Flows using Semi-Supervised Machine Learning

Improved Classification of Known and Unknown Network Traffic Flows using Semi-Supervised Machine Learning Improved Classification of Known and Unknown Network Traffic Flows using Semi-Supervised Machine Learning Timothy Glennan, Christopher Leckie, Sarah M. Erfani Department of Computing and Information Systems,

More information

TRANSPARENT OBJECT DETECTION USING REGIONS WITH CONVOLUTIONAL NEURAL NETWORK

TRANSPARENT OBJECT DETECTION USING REGIONS WITH CONVOLUTIONAL NEURAL NETWORK TRANSPARENT OBJECT DETECTION USING REGIONS WITH CONVOLUTIONAL NEURAL NETWORK 1 Po-Jen Lai ( 賴柏任 ), 2 Chiou-Shann Fuh ( 傅楸善 ) 1 Dept. of Electrical Engineering, National Taiwan University, Taiwan 2 Dept.

More information

Facial Expression Classification with Random Filters Feature Extraction

Facial Expression Classification with Random Filters Feature Extraction Facial Expression Classification with Random Filters Feature Extraction Mengye Ren Facial Monkey mren@cs.toronto.edu Zhi Hao Luo It s Me lzh@cs.toronto.edu I. ABSTRACT In our work, we attempted to tackle

More information

Volume 6, Issue 12, December 2018 International Journal of Advance Research in Computer Science and Management Studies

Volume 6, Issue 12, December 2018 International Journal of Advance Research in Computer Science and Management Studies ISSN: 2321-7782 (Online) e-isjn: A4372-3114 Impact Factor: 7.327 Volume 6, Issue 12, December 2018 International Journal of Advance Research in Computer Science and Management Studies Research Article

More information

Character Recognition

Character Recognition Character Recognition 5.1 INTRODUCTION Recognition is one of the important steps in image processing. There are different methods such as Histogram method, Hough transformation, Neural computing approaches

More information

EE368 Project Report CD Cover Recognition Using Modified SIFT Algorithm

EE368 Project Report CD Cover Recognition Using Modified SIFT Algorithm EE368 Project Report CD Cover Recognition Using Modified SIFT Algorithm Group 1: Mina A. Makar Stanford University mamakar@stanford.edu Abstract In this report, we investigate the application of the Scale-Invariant

More information

Storyline Reconstruction for Unordered Images

Storyline Reconstruction for Unordered Images Introduction: Storyline Reconstruction for Unordered Images Final Paper Sameedha Bairagi, Arpit Khandelwal, Venkatesh Raizaday Storyline reconstruction is a relatively new topic and has not been researched

More information

Invariant Recognition of Hand-Drawn Pictograms Using HMMs with a Rotating Feature Extraction

Invariant Recognition of Hand-Drawn Pictograms Using HMMs with a Rotating Feature Extraction Invariant Recognition of Hand-Drawn Pictograms Using HMMs with a Rotating Feature Extraction Stefan Müller, Gerhard Rigoll, Andreas Kosmala and Denis Mazurenok Department of Computer Science, Faculty of

More information

Final Report: Smart Trash Net: Waste Localization and Classification

Final Report: Smart Trash Net: Waste Localization and Classification Final Report: Smart Trash Net: Waste Localization and Classification Oluwasanya Awe oawe@stanford.edu Robel Mengistu robel@stanford.edu December 15, 2017 Vikram Sreedhar vsreed@stanford.edu Abstract Given

More information

An Exploration of Computer Vision Techniques for Bird Species Classification

An Exploration of Computer Vision Techniques for Bird Species Classification An Exploration of Computer Vision Techniques for Bird Species Classification Anne L. Alter, Karen M. Wang December 15, 2017 Abstract Bird classification, a fine-grained categorization task, is a complex

More information

Frequent Inner-Class Approach: A Semi-supervised Learning Technique for One-shot Learning

Frequent Inner-Class Approach: A Semi-supervised Learning Technique for One-shot Learning Frequent Inner-Class Approach: A Semi-supervised Learning Technique for One-shot Learning Izumi Suzuki, Koich Yamada, Muneyuki Unehara Nagaoka University of Technology, 1603-1, Kamitomioka Nagaoka, Niigata

More information

Spatial Localization and Detection. Lecture 8-1

Spatial Localization and Detection. Lecture 8-1 Lecture 8: Spatial Localization and Detection Lecture 8-1 Administrative - Project Proposals were due on Saturday Homework 2 due Friday 2/5 Homework 1 grades out this week Midterm will be in-class on Wednesday

More information

arxiv: v1 [cs.cv] 31 Mar 2016

arxiv: v1 [cs.cv] 31 Mar 2016 Object Boundary Guided Semantic Segmentation Qin Huang, Chunyang Xia, Wenchao Zheng, Yuhang Song, Hao Xu and C.-C. Jay Kuo arxiv:1603.09742v1 [cs.cv] 31 Mar 2016 University of Southern California Abstract.

More information

Table Identification and Information extraction in Spreadsheets

Table Identification and Information extraction in Spreadsheets Table Identification and Information extraction in Spreadsheets Elvis Koci 1,2, Maik Thiele 1, Oscar Romero 2, and Wolfgang Lehner 1 1 Technische Universität Dresden, Germany 2 Universitat Politècnica

More information

Realtime Object Recognition Using Decision Tree Learning

Realtime Object Recognition Using Decision Tree Learning Realtime Object Recognition Using Decision Tree Learning Dirk Wilking 1 and Thomas Röfer 2 1 Chair for Computer Science XI, Embedded Software Group, RWTH Aachen wilking@informatik.rwth-aachen.de 2 Center

More information

On the automatic classification of app reviews

On the automatic classification of app reviews Requirements Eng (2016) 21:311 331 DOI 10.1007/s00766-016-0251-9 RE 2015 On the automatic classification of app reviews Walid Maalej 1 Zijad Kurtanović 1 Hadeer Nabil 2 Christoph Stanik 1 Walid: please

More information

Structured Completion Predictors Applied to Image Segmentation

Structured Completion Predictors Applied to Image Segmentation Structured Completion Predictors Applied to Image Segmentation Dmitriy Brezhnev, Raphael-Joel Lim, Anirudh Venkatesh December 16, 2011 Abstract Multi-image segmentation makes use of global and local features

More information

AAUITEC at ImageCLEF 2015: Compound Figure Separation

AAUITEC at ImageCLEF 2015: Compound Figure Separation AAUITEC at ImageCLEF 2015: Compound Figure Separation Mario Taschwer 1 and Oge Marques 2 1 ITEC, Klagenfurt University (AAU), Austria, mario.taschwer@aau.at 2 Florida Atlantic University (FAU), Boca Raton,

More information

A Deep Learning Framework for Authorship Classification of Paintings

A Deep Learning Framework for Authorship Classification of Paintings A Deep Learning Framework for Authorship Classification of Paintings Kai-Lung Hua ( 花凱龍 ) Dept. of Computer Science and Information Engineering National Taiwan University of Science and Technology Taipei,

More information

Applied Statistics for Neuroscientists Part IIa: Machine Learning

Applied Statistics for Neuroscientists Part IIa: Machine Learning Applied Statistics for Neuroscientists Part IIa: Machine Learning Dr. Seyed-Ahmad Ahmadi 04.04.2017 16.11.2017 Outline Machine Learning Difference between statistics and machine learning Modeling the problem

More information

Efficient Voting Prediction for Pairwise Multilabel Classification

Efficient Voting Prediction for Pairwise Multilabel Classification Efficient Voting Prediction for Pairwise Multilabel Classification Eneldo Loza Mencía, Sang-Hyeun Park and Johannes Fürnkranz TU-Darmstadt - Knowledge Engineering Group Hochschulstr. 10 - Darmstadt - Germany

More information

String distance for automatic image classification

String distance for automatic image classification String distance for automatic image classification Nguyen Hong Thinh*, Le Vu Ha*, Barat Cecile** and Ducottet Christophe** *University of Engineering and Technology, Vietnam National University of HaNoi,

More information

Evaluation Measures. Sebastian Pölsterl. April 28, Computer Aided Medical Procedures Technische Universität München

Evaluation Measures. Sebastian Pölsterl. April 28, Computer Aided Medical Procedures Technische Universität München Evaluation Measures Sebastian Pölsterl Computer Aided Medical Procedures Technische Universität München April 28, 2015 Outline 1 Classification 1. Confusion Matrix 2. Receiver operating characteristics

More information

Weighted Convolutional Neural Network. Ensemble.

Weighted Convolutional Neural Network. Ensemble. Weighted Convolutional Neural Network Ensemble Xavier Frazão and Luís A. Alexandre Dept. of Informatics, Univ. Beira Interior and Instituto de Telecomunicações Covilhã, Portugal xavierfrazao@gmail.com

More information

Determining Document Skew Using Inter-Line Spaces

Determining Document Skew Using Inter-Line Spaces 2011 International Conference on Document Analysis and Recognition Determining Document Skew Using Inter-Line Spaces Boris Epshtein Google Inc. 1 1600 Amphitheatre Parkway, Mountain View, CA borisep@google.com

More information

An Empirical Study of Lazy Multilabel Classification Algorithms

An Empirical Study of Lazy Multilabel Classification Algorithms An Empirical Study of Lazy Multilabel Classification Algorithms E. Spyromitros and G. Tsoumakas and I. Vlahavas Department of Informatics, Aristotle University of Thessaloniki, 54124 Thessaloniki, Greece

More information

Using Machine Learning for Classification of Cancer Cells

Using Machine Learning for Classification of Cancer Cells Using Machine Learning for Classification of Cancer Cells Camille Biscarrat University of California, Berkeley I Introduction Cell screening is a commonly used technique in the development of new drugs.

More information

Polytechnic University of Tirana

Polytechnic University of Tirana 1 Polytechnic University of Tirana Department of Computer Engineering SIBORA THEODHOR ELINDA KAJO M ECE 2 Computer Vision OCR AND BEYOND THE PRESENTATION IS ORGANISED IN 3 PARTS : 3 Introduction, previous

More information

CIS UDEL Working Notes on ImageCLEF 2015: Compound figure detection task

CIS UDEL Working Notes on ImageCLEF 2015: Compound figure detection task CIS UDEL Working Notes on ImageCLEF 2015: Compound figure detection task Xiaolong Wang, Xiangying Jiang, Abhishek Kolagunda, Hagit Shatkay and Chandra Kambhamettu Department of Computer and Information

More information

Extraction of Semantic Text Portion Related to Anchor Link

Extraction of Semantic Text Portion Related to Anchor Link 1834 IEICE TRANS. INF. & SYST., VOL.E89 D, NO.6 JUNE 2006 PAPER Special Section on Human Communication II Extraction of Semantic Text Portion Related to Anchor Link Bui Quang HUNG a), Masanori OTSUBO,

More information

OPTICAL Character Recognition systems aim at converting

OPTICAL Character Recognition systems aim at converting ICDAR 2015 COMPETITION ON TEXT IMAGE SUPER-RESOLUTION 1 Boosting Optical Character Recognition: A Super-Resolution Approach Chao Dong, Ximei Zhu, Yubin Deng, Chen Change Loy, Member, IEEE, and Yu Qiao

More information

Learning to Match. Jun Xu, Zhengdong Lu, Tianqi Chen, Hang Li

Learning to Match. Jun Xu, Zhengdong Lu, Tianqi Chen, Hang Li Learning to Match Jun Xu, Zhengdong Lu, Tianqi Chen, Hang Li 1. Introduction The main tasks in many applications can be formalized as matching between heterogeneous objects, including search, recommendation,

More information

CRFVoter: Chemical Entity Mention, Gene and Protein Related Object recognition using a conglomerate of CRF based tools

CRFVoter: Chemical Entity Mention, Gene and Protein Related Object recognition using a conglomerate of CRF based tools CRFVoter: Chemical Entity Mention, Gene and Protein Related Object recognition using a conglomerate of CRF based tools Wahed Hemati, Alexander Mehler, and Tolga Uslu Text Technology Lab, Goethe Universitt

More information

Visual perception of unitary elements for layout analysis of unconstrained documents in heterogeneous databases

Visual perception of unitary elements for layout analysis of unconstrained documents in heterogeneous databases Visual perception of unitary elements for layout analysis of unconstrained documents in heterogeneous databases Baptiste Poirriez, Aurélie Lemaitre, Bertrand Coüasnon To cite this version: Baptiste Poirriez,

More information

Fine Classification of Unconstrained Handwritten Persian/Arabic Numerals by Removing Confusion amongst Similar Classes

Fine Classification of Unconstrained Handwritten Persian/Arabic Numerals by Removing Confusion amongst Similar Classes 2009 10th International Conference on Document Analysis and Recognition Fine Classification of Unconstrained Handwritten Persian/Arabic Numerals by Removing Confusion amongst Similar Classes Alireza Alaei

More information

Instant Camera Translation and Voicing for Signs

Instant Camera Translation and Voicing for Signs Instant Camera Translation and Voicing for Signs Kaifeng Chen Department of Applied Physics Stanford University Stanford, California 94305 Email:kfchen@stanford.edu Yao Zhou Department of Material Science

More information

Finger Print Enhancement Using Minutiae Based Algorithm

Finger Print Enhancement Using Minutiae Based Algorithm Available Online at www.ijcsmc.com International Journal of Computer Science and Mobile Computing A Monthly Journal of Computer Science and Information Technology IJCSMC, Vol. 3, Issue. 8, August 2014,

More information

Character Recognition from Google Street View Images

Character Recognition from Google Street View Images Character Recognition from Google Street View Images Indian Institute of Technology Course Project Report CS365A By Ritesh Kumar (11602) and Srikant Singh (12729) Under the guidance of Professor Amitabha

More information

Linear combinations of simple classifiers for the PASCAL challenge

Linear combinations of simple classifiers for the PASCAL challenge Linear combinations of simple classifiers for the PASCAL challenge Nik A. Melchior and David Lee 16 721 Advanced Perception The Robotics Institute Carnegie Mellon University Email: melchior@cmu.edu, dlee1@andrew.cmu.edu

More information

A Tagging Approach to Ontology Mapping

A Tagging Approach to Ontology Mapping A Tagging Approach to Ontology Mapping Colm Conroy 1, Declan O'Sullivan 1, Dave Lewis 1 1 Knowledge and Data Engineering Group, Trinity College Dublin {coconroy,declan.osullivan,dave.lewis}@cs.tcd.ie Abstract.

More information

An ICA based Approach for Complex Color Scene Text Binarization

An ICA based Approach for Complex Color Scene Text Binarization An ICA based Approach for Complex Color Scene Text Binarization Siddharth Kherada IIIT-Hyderabad, India siddharth.kherada@research.iiit.ac.in Anoop M. Namboodiri IIIT-Hyderabad, India anoop@iiit.ac.in

More information

Lecture 5: Object Detection

Lecture 5: Object Detection Object Detection CSED703R: Deep Learning for Visual Recognition (2017F) Lecture 5: Object Detection Bohyung Han Computer Vision Lab. bhhan@postech.ac.kr 2 Traditional Object Detection Algorithms Region-based

More information

On-line handwriting recognition using Chain Code representation

On-line handwriting recognition using Chain Code representation On-line handwriting recognition using Chain Code representation Final project by Michal Shemesh shemeshm at cs dot bgu dot ac dot il Introduction Background When one preparing a first draft, concentrating

More information

Indoor Object Recognition of 3D Kinect Dataset with RNNs

Indoor Object Recognition of 3D Kinect Dataset with RNNs Indoor Object Recognition of 3D Kinect Dataset with RNNs Thiraphat Charoensripongsa, Yue Chen, Brian Cheng 1. Introduction Recent work at Stanford in the area of scene understanding has involved using

More information

Investigation of Machine Learning Algorithm Compared to Fuzzy Logic in Wild Fire Smoke Detection Applications

Investigation of Machine Learning Algorithm Compared to Fuzzy Logic in Wild Fire Smoke Detection Applications Investigation of Machine Learning Algorithm Compared to Fuzzy Logic in Wild Fire Smoke Detection Applications Azarm Nowzad, Andreas Jock, Klaus Jäckel IQ wireless GmbH, Berlin, Germany Azarm Nowzad, Ralf

More information

Tri-modal Human Body Segmentation

Tri-modal Human Body Segmentation Tri-modal Human Body Segmentation Master of Science Thesis Cristina Palmero Cantariño Advisor: Sergio Escalera Guerrero February 6, 2014 Outline 1 Introduction 2 Tri-modal dataset 3 Proposed baseline 4

More information

Visual object classification by sparse convolutional neural networks

Visual object classification by sparse convolutional neural networks Visual object classification by sparse convolutional neural networks Alexander Gepperth 1 1- Ruhr-Universität Bochum - Institute for Neural Dynamics Universitätsstraße 150, 44801 Bochum - Germany Abstract.

More information

Efficient Segmentation-Aided Text Detection For Intelligent Robots

Efficient Segmentation-Aided Text Detection For Intelligent Robots Efficient Segmentation-Aided Text Detection For Intelligent Robots Junting Zhang, Yuewei Na, Siyang Li, C.-C. Jay Kuo University of Southern California Outline Problem Definition and Motivation Related

More information

DEEP STRUCTURED OUTPUT LEARNING FOR UNCONSTRAINED TEXT RECOGNITION

DEEP STRUCTURED OUTPUT LEARNING FOR UNCONSTRAINED TEXT RECOGNITION DEEP STRUCTURED OUTPUT LEARNING FOR UNCONSTRAINED TEXT RECOGNITION Max Jaderberg, Karen Simonyan, Andrea Vedaldi, Andrew Zisserman Visual Geometry Group, Department Engineering Science, University of Oxford,

More information

Empirical Analysis of Single and Multi Document Summarization using Clustering Algorithms

Empirical Analysis of Single and Multi Document Summarization using Clustering Algorithms Engineering, Technology & Applied Science Research Vol. 8, No. 1, 2018, 2562-2567 2562 Empirical Analysis of Single and Multi Document Summarization using Clustering Algorithms Mrunal S. Bewoor Department

More information

OTCYMIST: Otsu-Canny Minimal Spanning Tree for Born-Digital Images

OTCYMIST: Otsu-Canny Minimal Spanning Tree for Born-Digital Images OTCYMIST: Otsu-Canny Minimal Spanning Tree for Born-Digital Images Deepak Kumar and A G Ramakrishnan Medical Intelligence and Language Engineering Laboratory Department of Electrical Engineering, Indian

More information

Application of Additive Groves Ensemble with Multiple Counts Feature Evaluation to KDD Cup 09 Small Data Set

Application of Additive Groves Ensemble with Multiple Counts Feature Evaluation to KDD Cup 09 Small Data Set Application of Additive Groves Application of Additive Groves Ensemble with Multiple Counts Feature Evaluation to KDD Cup 09 Small Data Set Daria Sorokina Carnegie Mellon University Pittsburgh PA 15213

More information

Classifying a specific image region using convolutional nets with an ROI mask as input

Classifying a specific image region using convolutional nets with an ROI mask as input Classifying a specific image region using convolutional nets with an ROI mask as input 1 Sagi Eppel Abstract Convolutional neural nets (CNN) are the leading computer vision method for classifying images.

More information

CMU Lecture 18: Deep learning and Vision: Convolutional neural networks. Teacher: Gianni A. Di Caro

CMU Lecture 18: Deep learning and Vision: Convolutional neural networks. Teacher: Gianni A. Di Caro CMU 15-781 Lecture 18: Deep learning and Vision: Convolutional neural networks Teacher: Gianni A. Di Caro DEEP, SHALLOW, CONNECTED, SPARSE? Fully connected multi-layer feed-forward perceptrons: More powerful

More information

Applying the KISS Principle for the CLEF- IP 2010 Prior Art Candidate Patent Search Task

Applying the KISS Principle for the CLEF- IP 2010 Prior Art Candidate Patent Search Task Applying the KISS Principle for the CLEF- IP 2010 Prior Art Candidate Patent Search Task Walid Magdy, Gareth J.F. Jones Centre for Next Generation Localisation School of Computing Dublin City University,

More information

A THREE LAYERED MODEL TO PERFORM CHARACTER RECOGNITION FOR NOISY IMAGES

A THREE LAYERED MODEL TO PERFORM CHARACTER RECOGNITION FOR NOISY IMAGES INTERNATIONAL JOURNAL OF RESEARCH IN COMPUTER APPLICATIONSAND ROBOTICS ISSN 2320-7345 A THREE LAYERED MODEL TO PERFORM CHARACTER RECOGNITION FOR NOISY IMAGES 1 Neha, 2 Anil Saroliya, 3 Varun Sharma 1,

More information

New Binarization Approach Based on Text Block Extraction

New Binarization Approach Based on Text Block Extraction 2011 International Conference on Document Analysis and Recognition New Binarization Approach Based on Text Block Extraction Ines Ben Messaoud, Hamid Amiri Laboratoire des Systèmes et Traitement de Signal

More information

ECE 285 Class Project Report

ECE 285 Class Project Report ECE 285 Class Project Report Based on Source localization in an ocean waveguide using supervised machine learning Yiwen Gong ( yig122@eng.ucsd.edu), Yu Chai( yuc385@eng.ucsd.edu ), Yifeng Bu( ybu@eng.ucsd.edu

More information

CLEF-IP 2009: Exploring Standard IR Techniques on Patent Retrieval

CLEF-IP 2009: Exploring Standard IR Techniques on Patent Retrieval DCU @ CLEF-IP 2009: Exploring Standard IR Techniques on Patent Retrieval Walid Magdy, Johannes Leveling, Gareth J.F. Jones Centre for Next Generation Localization School of Computing Dublin City University,

More information

Object Detection Based on Deep Learning

Object Detection Based on Deep Learning Object Detection Based on Deep Learning Yurii Pashchenko AI Ukraine 2016, Kharkiv, 2016 Image classification (mostly what you ve seen) http://tutorial.caffe.berkeleyvision.org/caffe-cvpr15-detection.pdf

More information

A FRAMEWORK OF EXTRACTING MULTI-SCALE FEATURES USING MULTIPLE CONVOLUTIONAL NEURAL NETWORKS. Kuan-Chuan Peng and Tsuhan Chen

A FRAMEWORK OF EXTRACTING MULTI-SCALE FEATURES USING MULTIPLE CONVOLUTIONAL NEURAL NETWORKS. Kuan-Chuan Peng and Tsuhan Chen A FRAMEWORK OF EXTRACTING MULTI-SCALE FEATURES USING MULTIPLE CONVOLUTIONAL NEURAL NETWORKS Kuan-Chuan Peng and Tsuhan Chen School of Electrical and Computer Engineering, Cornell University, Ithaca, NY

More information

arxiv: v1 [cs.cv] 20 Dec 2016

arxiv: v1 [cs.cv] 20 Dec 2016 End-to-End Pedestrian Collision Warning System based on a Convolutional Neural Network with Semantic Segmentation arxiv:1612.06558v1 [cs.cv] 20 Dec 2016 Heechul Jung heechul@dgist.ac.kr Min-Kook Choi mkchoi@dgist.ac.kr

More information

Visual Appearance based Document Classification Methods: Performance Evaluation and Benchmarking

Visual Appearance based Document Classification Methods: Performance Evaluation and Benchmarking Visual Appearance based Document Classification Methods: Performance Evaluation and Benchmarking Syed Saqib Bukhari, Andreas Dengel German Research Center for Artificial Intelligence (DFKI) http://www.dfki.de/

More information

Text Recognition in Videos using a Recurrent Connectionist Approach

Text Recognition in Videos using a Recurrent Connectionist Approach Author manuscript, published in "ICANN - 22th International Conference on Artificial Neural Networks, Lausanne : Switzerland (2012)" DOI : 10.1007/978-3-642-33266-1_22 Text Recognition in Videos using

More information

Network Traffic Measurements and Analysis

Network Traffic Measurements and Analysis DEIB - Politecnico di Milano Fall, 2017 Sources Hastie, Tibshirani, Friedman: The Elements of Statistical Learning James, Witten, Hastie, Tibshirani: An Introduction to Statistical Learning Andrew Ng:

More information

Handwritten Script Recognition at Block Level

Handwritten Script Recognition at Block Level Chapter 4 Handwritten Script Recognition at Block Level -------------------------------------------------------------------------------------------------------------------------- Optical character recognition

More information

Data mining with Support Vector Machine

Data mining with Support Vector Machine Data mining with Support Vector Machine Ms. Arti Patle IES, IPS Academy Indore (M.P.) artipatle@gmail.com Mr. Deepak Singh Chouhan IES, IPS Academy Indore (M.P.) deepak.schouhan@yahoo.com Abstract: Machine

More information

A modified and fast Perceptron learning rule and its use for Tag Recommendations in Social Bookmarking Systems

A modified and fast Perceptron learning rule and its use for Tag Recommendations in Social Bookmarking Systems A modified and fast Perceptron learning rule and its use for Tag Recommendations in Social Bookmarking Systems Anestis Gkanogiannis and Theodore Kalamboukis Department of Informatics Athens University

More information

Business Forms Classification using Earth Mover s Distance

Business Forms Classification using Earth Mover s Distance Business Forms Classification using Earth Mover s Distance Abstract Form Classification has not been focused on for the last decade. Unfortunately the algorithms published mainly in the 80s and 90s do

More information

Boosting Simple Model Selection Cross Validation Regularization. October 3 rd, 2007 Carlos Guestrin [Schapire, 1989]

Boosting Simple Model Selection Cross Validation Regularization. October 3 rd, 2007 Carlos Guestrin [Schapire, 1989] Boosting Simple Model Selection Cross Validation Regularization Machine Learning 10701/15781 Carlos Guestrin Carnegie Mellon University October 3 rd, 2007 1 Boosting [Schapire, 1989] Idea: given a weak

More information

ARE you? CS229 Final Project. Jim Hefner & Roddy Lindsay

ARE you? CS229 Final Project. Jim Hefner & Roddy Lindsay ARE you? CS229 Final Project Jim Hefner & Roddy Lindsay 1. Introduction We use machine learning algorithms to predict attractiveness ratings for photos. There is a wealth of psychological evidence indicating

More information

Object Purpose Based Grasping

Object Purpose Based Grasping Object Purpose Based Grasping Song Cao, Jijie Zhao Abstract Objects often have multiple purposes, and the way humans grasp a certain object may vary based on the different intended purposes. To enable

More information

N.Priya. Keywords Compass mask, Threshold, Morphological Operators, Statistical Measures, Text extraction

N.Priya. Keywords Compass mask, Threshold, Morphological Operators, Statistical Measures, Text extraction Volume, Issue 8, August ISSN: 77 8X International Journal of Advanced Research in Computer Science and Software Engineering Research Paper Available online at: www.ijarcsse.com A Combined Edge-Based Text

More information

Classification of objects from Video Data (Group 30)

Classification of objects from Video Data (Group 30) Classification of objects from Video Data (Group 30) Sheallika Singh 12665 Vibhuti Mahajan 12792 Aahitagni Mukherjee 12001 M Arvind 12385 1 Motivation Video surveillance has been employed for a long time

More information

anyocr: An Open-Source OCR System for Historical Archives

anyocr: An Open-Source OCR System for Historical Archives anyocr: An Open-Source OCR System for Historical Archives Abstract Currently an intensive amount of research is going on in the field of digitizing historical Archives for converting scanned page images

More information

3 Object Detection. BVM 2018 Tutorial: Advanced Deep Learning Methods. Paul F. Jaeger, Division of Medical Image Computing

3 Object Detection. BVM 2018 Tutorial: Advanced Deep Learning Methods. Paul F. Jaeger, Division of Medical Image Computing 3 Object Detection BVM 2018 Tutorial: Advanced Deep Learning Methods Paul F. Jaeger, of Medical Image Computing What is object detection? classification segmentation obj. detection (1 label per pixel)

More information

Semi supervised clustering for Text Clustering

Semi supervised clustering for Text Clustering Semi supervised clustering for Text Clustering N.Saranya 1 Assistant Professor, Department of Computer Science and Engineering, Sri Eshwar College of Engineering, Coimbatore 1 ABSTRACT: Based on clustering

More information

LSTM: An Image Classification Model Based on Fashion-MNIST Dataset

LSTM: An Image Classification Model Based on Fashion-MNIST Dataset LSTM: An Image Classification Model Based on Fashion-MNIST Dataset Kexin Zhang, Research School of Computer Science, Australian National University Kexin Zhang, U6342657@anu.edu.au Abstract. The application

More information

A Study to Recognize Printed Gujarati Characters Using Tesseract OCR

A Study to Recognize Printed Gujarati Characters Using Tesseract OCR A Study to Recognize Printed Gujarati Characters Using Tesseract OCR Milind Kumar Audichya 1, Jatinderkumar R. Saini 2 1, 2 Computer Science, Gujarat Technological University Abstract: Optical Character

More information

DEEP BLIND IMAGE QUALITY ASSESSMENT

DEEP BLIND IMAGE QUALITY ASSESSMENT DEEP BLIND IMAGE QUALITY ASSESSMENT BY LEARNING SENSITIVITY MAP Jongyoo Kim, Woojae Kim and Sanghoon Lee ICASSP 2018 Deep Learning and Convolutional Neural Networks (CNNs) SOTA in computer vision & image

More information

Bone Age Classification Using the Discriminative Generalized Hough Transform

Bone Age Classification Using the Discriminative Generalized Hough Transform Bone Age Classification Using the Discriminative Generalized Hough Transform Markus Brunk 1, Heike Ruppertshofen 1,2, Sarah Schmidt 2,3, Peter Beyerlein 3, Hauke Schramm 1 1 Institute of Applied Computer

More information

Deep Learning for Object detection & localization

Deep Learning for Object detection & localization Deep Learning for Object detection & localization RCNN, Fast RCNN, Faster RCNN, YOLO, GAP, CAM, MSROI Aaditya Prakash Sep 25, 2018 Image classification Image classification Whole of image is classified

More information

Mask R-CNN. presented by Jiageng Zhang, Jingyao Zhan, Yunhan Ma

Mask R-CNN. presented by Jiageng Zhang, Jingyao Zhan, Yunhan Ma Mask R-CNN presented by Jiageng Zhang, Jingyao Zhan, Yunhan Ma Mask R-CNN Background Related Work Architecture Experiment Mask R-CNN Background Related Work Architecture Experiment Background From left

More information

Depth from Stereo. Dominic Cheng February 7, 2018

Depth from Stereo. Dominic Cheng February 7, 2018 Depth from Stereo Dominic Cheng February 7, 2018 Agenda 1. Introduction to stereo 2. Efficient Deep Learning for Stereo Matching (W. Luo, A. Schwing, and R. Urtasun. In CVPR 2016.) 3. Cascade Residual

More information

Stat 342 Exam 3 Fall 2014

Stat 342 Exam 3 Fall 2014 Stat 34 Exam 3 Fall 04 I have neither given nor received unauthorized assistance on this exam. Name Signed Date Name Printed There are questions on the following 6 pages. Do as many of them as you can

More information

Combining Neural Networks and Log-linear Models to Improve Relation Extraction

Combining Neural Networks and Log-linear Models to Improve Relation Extraction Combining Neural Networks and Log-linear Models to Improve Relation Extraction Thien Huu Nguyen and Ralph Grishman Computer Science Department, New York University {thien,grishman}@cs.nyu.edu Outline Relation

More information

Kaggle Data Science Bowl 2017 Technical Report

Kaggle Data Science Bowl 2017 Technical Report Kaggle Data Science Bowl 2017 Technical Report qfpxfd Team May 11, 2017 1 Team Members Table 1: Team members Name E-Mail University Jia Ding dingjia@pku.edu.cn Peking University, Beijing, China Aoxue Li

More information

ViTraM: VIsualization of TRAnscriptional Modules

ViTraM: VIsualization of TRAnscriptional Modules ViTraM: VIsualization of TRAnscriptional Modules Version 2.0 October 1st, 2009 KULeuven, Belgium 1 Contents 1 INTRODUCTION AND INSTALLATION... 4 1.1 Introduction...4 1.2 Software structure...5 1.3 Requirements...5

More information

Face Quality Assessment System in Video Sequences

Face Quality Assessment System in Video Sequences Face Quality Assessment System in Video Sequences Kamal Nasrollahi, Thomas B. Moeslund Laboratory of Computer Vision and Media Technology, Aalborg University Niels Jernes Vej 14, 9220 Aalborg Øst, Denmark

More information

Multimodal Medical Image Retrieval based on Latent Topic Modeling

Multimodal Medical Image Retrieval based on Latent Topic Modeling Multimodal Medical Image Retrieval based on Latent Topic Modeling Mandikal Vikram 15it217.vikram@nitk.edu.in Suhas BS 15it110.suhas@nitk.edu.in Aditya Anantharaman 15it201.aditya.a@nitk.edu.in Sowmya Kamath

More information

Dynamic Optimization of Generalized SQL Queries with Horizontal Aggregations Using K-Means Clustering

Dynamic Optimization of Generalized SQL Queries with Horizontal Aggregations Using K-Means Clustering Dynamic Optimization of Generalized SQL Queries with Horizontal Aggregations Using K-Means Clustering Abstract Mrs. C. Poongodi 1, Ms. R. Kalaivani 2 1 PG Student, 2 Assistant Professor, Department of

More information

SOMSN: An Effective Self Organizing Map for Clustering of Social Networks

SOMSN: An Effective Self Organizing Map for Clustering of Social Networks SOMSN: An Effective Self Organizing Map for Clustering of Social Networks Fatemeh Ghaemmaghami Research Scholar, CSE and IT Dept. Shiraz University, Shiraz, Iran Reza Manouchehri Sarhadi Research Scholar,

More information

Deep Learning in Visual Recognition. Thanks Da Zhang for the slides

Deep Learning in Visual Recognition. Thanks Da Zhang for the slides Deep Learning in Visual Recognition Thanks Da Zhang for the slides Deep Learning is Everywhere 2 Roadmap Introduction Convolutional Neural Network Application Image Classification Object Detection Object

More information

Classification/Regression Trees and Random Forests

Classification/Regression Trees and Random Forests Classification/Regression Trees and Random Forests Fabio G. Cozman - fgcozman@usp.br November 6, 2018 Classification tree Consider binary class variable Y and features X 1,..., X n. Decide Ŷ after a series

More information

Layout Segmentation of Scanned Newspaper Documents

Layout Segmentation of Scanned Newspaper Documents , pp-05-10 Layout Segmentation of Scanned Newspaper Documents A.Bandyopadhyay, A. Ganguly and U.Pal CVPR Unit, Indian Statistical Institute 203 B T Road, Kolkata, India. Abstract: Layout segmentation algorithms

More information

Comparison of various classification models for making financial decisions

Comparison of various classification models for making financial decisions Comparison of various classification models for making financial decisions Vaibhav Mohan Computer Science Department Johns Hopkins University Baltimore, MD 21218, USA vmohan3@jhu.edu Abstract Banks are

More information

Forward Feature Selection Using Residual Mutual Information

Forward Feature Selection Using Residual Mutual Information Forward Feature Selection Using Residual Mutual Information Erik Schaffernicht, Christoph Möller, Klaus Debes and Horst-Michael Gross Ilmenau University of Technology - Neuroinformatics and Cognitive Robotics

More information

Yiqi Yan. May 10, 2017

Yiqi Yan. May 10, 2017 Yiqi Yan May 10, 2017 P a r t I F u n d a m e n t a l B a c k g r o u n d s Convolution Single Filter Multiple Filters 3 Convolution: case study, 2 filters 4 Convolution: receptive field receptive field

More information

End-To-End Spam Classification With Neural Networks

End-To-End Spam Classification With Neural Networks End-To-End Spam Classification With Neural Networks Christopher Lennan, Bastian Naber, Jan Reher, Leon Weber 1 Introduction A few years ago, the majority of the internet s network traffic was due to spam

More information