Haar-features training parameters analysis in boosting based machine learning for improved face detection

Abstract

Haar features have been used in most of the works in literature as key components in the task of object as well as face detection. Training of Haar features is an important step in the development of overall machine learning based face detection system. In this work, we have done investigation in the variations of a number of training parameters during AdaBoost based machine learning of Haar features with respect to size of training images. A number of observations have been drawn based on the variations of these parameters during training process. Training parameters True Detection Rate and Figure of Merit have been used as weighing parameter in formation of detection cascade for improvement in original AdaBoost framework used for machine learning of Haar features. Statistical analysis based on variations of training parameters has been done for selection of efficient learners from a large pool of available features which further are cascaded to make the strong classifiers.we have been able to achieve the maximum percentage reduction in training time of 47.20 corresponding to training images size 25x25 pixels with the help of statistical screening of Haar features prior to AdaBoost. The proposed system has been developed using training dataset from CBCL and successfully tested on facial images from datasets WIDER FACE a Face Detection Benchmark dataset, Annotated Faces in the Wild (AFW) and Face Detection Data Set and Benchmark (FDDB). and been able to achieve promising results in terms of training time computedin range of 0.31-2.83hrs, maximum detection rate 96.21% and minimum detection time16.95ms.

Intelligent face sketch recognition system using shearlet transform and convolutional neural network model

“…Before delving into this section, it is essential to recall that face detection is a fundamental step in the recognition pipeline, as it involves locating and extracting facial regions from an image or sketch [8]. Also, the use of convolutional neural networks (CNNs) [911] in image processing has led to significant improvement and rapid advancement in applications related to this field [12,13,9].…”

“…Directions: We employ four directions (4,8,16,16) for the shearlet transform. These directions facilitate the extraction of directional information from the input images, which is important in distinguishing facial characteristics and structures.…”

Computer vision and deep learning for fish classification in underwater habitats: A survey

“…Although there are many existing models, most of the classical non-DL models are based on local and engineered features. These include works using Haar features (Mutneja & Singh, 2021), scale-invariant feature transform (SIFT) (Lindeberg, 2012) and histogram of oriented gradient (HOG) (Dalal & Triggs, 2005), which need hand-engineered algorithms. Because these algorithms are not suitable for recognizing images of untrained animals and cannot capture fish features from complex backgrounds, they usually use a large number of manually extracted samples to build classifiers.…”