Bio: Aba Graf is an academic researcher from Max Planck Society. The author has contributed to research in topics: Support vector machine & Liquid crystal. The author has an hindex of 2, co-authored 5 publications receiving 12 citations.
01 Aug 2002
TL;DR: A novel representation for image classification is proposed which exploits the temporal information inherent in natural visual input and shows that the feature representation significantly outperforms the view representation.
Abstract: In this paper a novel representation for image classification is proposed which exploits the temporal information inherent in natural visual input. Image sequences are represented by a set of salient features which are found by tracking of visual features. In the context of a multi-class classification problem this representation is compared against a representation using only raw image data. The dataset consists of image sequences generated from a processed version of the MPI face database. We consider two types of multi-class SVMs and benchmark them against nearest-neighbor classifiers. By introducing a new set of SVM kernel functions we show that the feature representation significantly outperforms the view representation.
01 Feb 2002
01 Jan 2002
01 Mar 2004
01 Feb 2003
01 Jan 1999
TL;DR: In this article, heat transfer is used for measurement techniques in Turbomachinery and GTT is used to transfer heat from a turbine to a turbine in the form of heat transfer.
Abstract: Keywords: Heat Transfer ; Measurement Techniques ; Turbomachinery ; GTT ; LTT Reference LTT-ARTICLE-2003-002doi:10.1115/1.1578501View record in Web of Science Record created on 2007-04-18, modified on 2017-05-10
••02 Dec 2013
TL;DR: The proposed camera based system can provide advantages over the traditional radar/laser based warning systems, in which both the LDW and the FCW information cannot be provided with the same system, and provides additional flexibility at a lower cost.
Abstract: This paper presents a unique camera-based Forward Collision Warning (FCW) and Lane Departure Warning (LDW) system to improve the safety of road vehicle transportation The video used in the algorithm is captured by an in-car camera Initially, a Support Vector Machine (SVM) classifier is applied to the first frame of the video to locate the moving vehicle of interest in front of the host vehicle Following this step, two separate warning systems, namely FCW and LDW are designed For the FCW, the Time to Collision (TTC) is determined through the scale change method, and the FCW system will be activated when TTC is less than a predefined threshold value For the LDW system, the lane position information is analyzed and the warning is triggered if there is a lane departure without the use of blinkers The proposed camera based system can provide advantages over the traditional radar/laser based warning systems, in which both the LDW and the FCW information cannot be provided with the same system Furthermore, the proposed system provides additional flexibility at a lower cost
••06 May 2012
TL;DR: The experimental results show that the proposed pavement segmentation and crack detection system is robust and can effectively be used in pavement images with complicated background components such as trees, houses, etc.
Abstract: This paper presents a pavement segmentation and crack detection system from pavement images with complicated background information. The proposed method consists of three steps. In the first step, a Support Vector Machine, which shows a high degree of accuracy in classifying data, was employed to classify the image into two categories: a pavement group and a background group. In the second step, the crack was extracted by a fractal thresholding. Finally, a Radon Transform was applied to the crack image to classify the cracks into four different types. The experimental results show that the proposed system is robust and can effectively be used in pavement images with complicated background components such as trees, houses, etc.
••21 Mar 2012
TL;DR: Self Training Algorithm is used for Semi Supervised Learning and Winner-Takes-All strategy is applied to perform Multi Class Classification, which provides Classification accuracy up to 93%.
Abstract: It is always important in the Software Industry to know about what types of bugs are getting reported into the applications developed or maintained by them. Categorizing bugs based on their characteristics helps Software Development team to take appropriate actions in order to reduce similar defects that might get reported in future releases. Defects or Bugs can be classified into many classes, for which a training set is required, known as the Class Label Data Set. If Classification is performed manually then it will consume more time and efforts. Also, human resource having expert testing skills & domain knowledge will be required for labelling the data. Therefore Semi Supervised Techniques are been used to reduce the work of labelling dataset, which takes some labeled with unlabeled dataset to train the classifier. In this paper Self Training Algorithm is used for Semi Supervised Learning and Winner-Takes-All strategy is applied to perform Multi Class Classification. This model provides Classification accuracy up to 93%.