Machine Learning Methods for Visual Object Detection.

      Аннотация: This book presents practical methods for detecting common object classes such as people, cars, cows, chairs, etc., in real world images. In particular, it discusses a range of visual representations (Histograms of Oriented Gradients (HOG), Local Binary Patterns (LBP), Local Ternary Patterns (LTP), Local Quantized Patterns (LQP) and similar techniques), dimensionality reduction (Partial Least Squares and SVM Weight Sparsification) and learning methods (Latent and Non-Latent Support Vector Machines) for the problem of object detection. These methods are presented from a practical perspective and shown to give state-of-the-art performance on a range of challenging public datasets.

Ключевые слова: machine learning, object detection, Visual recognition, Histogram of Oriented Gradients (HOG), Local Binary Patterns (LBP), Local Ternary Patterns (LTP), Local Quantized Patterns (LQP), Dimensionality reduction, features, Visual descriptors, Support Vector Machines (SVM), partial least squares (PLS), Pascal Datasets, PASCAL Visual Object Classes (VOC) Challenge, Weight truncation, INRIA Person dataset, Human/Pedestrian detection