A Cue - Integration Scheme for Object Recognition Using Discriminative Accumulation

نویسنده

  • Maria-Elena Nilsback
چکیده

Robustness is an important property for any object recognition system. A way to increase robustness is by using multiple cues. This thesis proposes a cue-integration scheme for discriminative classifiers. The basic idea is to use one classifier for each cue and combine the outputs of the classifiers. The outputs are combined through weighted summation, which allows different cues to have different influences on the classification. The cue-integration scheme is derived and thoroughly tested. Benchmarking against other methods using multiple cues, a voting scheme and a probabilistic accumulation scheme, shows very good performance for the new cue-integration scheme. However, there is evidence that it might suffer from generalization problems as the number of object classes increases. Based on the last fact a decision tree using the new cue-integration scheme for each decision is proposed. This is also thoroughly tested. As expected, this method does not suffer from generalization problems and benchmarking gives evidence of an overall improved performance. Kombinering av ledtrådar för objektigenkänning genom diskriminativ ackumulering Sammanfattning Det är viktigt för objektigenkännings-system att vara robusta mot oförutsedda variationer i bilder. Ett sätt att göra objektigenkänningen mer robust är att använda flera ledtrådar. Denna rapport presenterar en metod för att kombinera olika ledtrådar för att på så sätt uppnå mer robusta system. Grundidén är att använda en diskriminativ klassificerare för varje ledtråd och sedan kombinera utdata från dessa klassificerare. Utdata kombineras genom viktad addition, vilket innebär att olika ledtrådar har olika inverkan vid klassificeringen. Metoden härleds och testas noggrant. Jämförelse med beslutsträd och probabilistisk ackumulering visar mycket goda resultat för den nya metoden, även om det visar sig att den nya metoden har vissa generaliseringsproblem då antalet klasser ökar. En metod inspirerad av det sista påståendet föreslås, där ackumuleringsmetoden kombineras med ett beslutsträd så att varje binärt beslut görs med hjälp av ackumulering. Denna metod testas också noggrant och det visar sig som förväntat att jämfört med de andra metoderna uppnås genomgående förbättrade resultat.

برای دانلود متن کامل این مقاله و بیش از 32 میلیون مقاله دیگر ابتدا ثبت نام کنید

ثبت نام

اگر عضو سایت هستید لطفا وارد حساب کاربری خود شوید

منابع مشابه

Discriminative cue integration for medical image annotation

Automatic annotation of medical images is an increasingly important tool for physicians in their daily activity. Hospitals produce nowadays an increasing amount of data. Manual annotation is very costly and prone to human mistakes. This paper proposes a multi-cue approach to automatic medical image annotation. We represent images using global and local features. These cues are then combined tog...

متن کامل

Integration of Regions and Contours for Object Recognition

We present an integrated approach combining region and contour–based techniques to enhance both segmentation and recognition processes. This cue integration operates on the level of contour–based groups and complete regions, which are matched to reflect a common cause in the image (and thus in the scene). Additionally, we realize a top–down scheme controlling the segmentation process on the bas...

متن کامل

Biologically Motivated Audio-Visual Cue Integration for Object Categorization

Auditory and visual cues are important sensor inputs for biological and artificial systems. They provide crucial information for navigating environments, recognizing categories, animals and people. How to combine effectively these two sensory channels is still an open issue. As a step towards this goal, this paper presents a comparison between three different multi-modal integration strategies,...

متن کامل

CLEF2007: Image Annotation Task: an SVM-based Cue Integration Approach

This paper presents the algorithms and results of our participation to the medical image annotation task of ImageCLEFmed 2007. We proposed, as a general strategy, a multi-cue approach where images are represented both by global and local descriptors, so to capture different types of information. These cues are combined during the classification step following two alternative SVM-based strategie...

متن کامل

Probabilistic Combination of Visual Cues for Object Classification

Recent solutions to object classification have focused on the decomposition of objects into representative parts. However, the vast majority of these methods are based on single visual cue measurements. Psychophysical evidence suggests that humans use multiple visual cues to accomplish recognition. In this paper, we address the problem of integrating multiple visual information for object recog...

متن کامل

ذخیره در منابع من


  با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید

برای دانلود متن کامل این مقاله و بیش از 32 میلیون مقاله دیگر ابتدا ثبت نام کنید

ثبت نام

اگر عضو سایت هستید لطفا وارد حساب کاربری خود شوید

عنوان ژورنال:

دوره   شماره 

صفحات  -

تاریخ انتشار 2004