K. Sudhakar

Bio: K. Sudhakar is an academic researcher from VIT University. The author has contributed to research in topics: Pattern recognition (psychology) & Gabor wavelet. The author has an hindex of 2, co-authored 2 publications receiving 14 citations.

An Accurate Facial Component Detection Using Gabor Filter

Abstract: Face detection is a critical task to be resolved in a variety of applications. Since faces include various expressions it becomes a difficult task to detect the exact output. Face detection not only play a main role in personal identification but also in various fields which includes but not limited to image processing, pattern recognition, graphics and other application areas. The proposed system performs the face detection and facial components using Gabor filter. The results show accurate detection of facial components

Facial identification of twins based on fusion score method

Abstract: Face recognition system contains lots of challenges due to various environmental factors, background variations, poor quality of camera, different illumination and others. Since twins are involved with criminal activities, twin identification becomes an essential task. The proposed system is focused on identifying the identical twins for the still images. The fusion based approach has been implemented in the proposed system. It combines the features extracted by using Principal Component Analysis (PCA), Histogram Oriented of Gradients (HOG), Local Binary Pattern (LBP), Gabor and distance between the facial components. Three types of fusion such as Decision Level Fusion, Feature Level Fusion and Score Level Fusion are used in the proposed approach. Based on these fusion generated scores, the twin has been identified. In the proposed system, Particle Swarm Optimization is used for the best feature selection and SVM classifier is used for training and testing the image. The proposed system provides better results when compared with the other twin detection techniques.

Web User Profile Generation and Discovery Analysis using LSTM Architecture

Abstract: In today's technology-driven world, a user profile is a virtual representation of each user, containing various user information such as personal, interest and preference data. These profiles are the result of a user profiling process and are essential to personalizing the service. As the amount of information available on the Internet increases and the number of different users, customization becomes a priority. Due to the large amount of information available on the Internet, referral systems that aim to provide relevant information to users are becoming increasingly important and popular. Various methods, methodologies and algorithms have been proposed in the literature for the user analysis process. Creating automated user profiles is a big challenge in creating adaptive customized applications. In this work proposed the method, Long Short-Term Architecture (LSTM) is User profile is an important issue for both information and service customization. Based on the original information, the user's topic preference and text emotional features into attention information and combines various formats and LSTM (Long Short Term Memory) models to describe and predict the elements of informal community clients. At last, the trial consequences of different gatherings show that the concern-based LSTM model proposed can accomplish improved results than the right now regularly involved strategies in recognizing client character qualities, and the model has great speculation, which implies that it has this capacity.

Facial Component-Landmark Detection With Weakly-Supervised LR-CNN

Abstract: In this paper, we propose a weakly supervised landmark-region-based convolutional neural network (LR-CNN) framework to detect facial component and landmark simultaneously. Most of the existing course-to-fine facial detectors fail to detect landmark accurately without lots of fully labeled data, which are costly to obtain. We can handle the task with a small amount of finely labeled data. First, deep convolutional generative adversarial networks are utilized to generate training samples with weak labels, as data preparation. Then, through weakly supervised learning, our LR-CNN model can be trained effectively with a small amount of finely labeled data and a large amount of generated weakly labeled data. Notably, our approach can handle the situation when large occlusion areas occur, as we localize visible facial components before predicting corresponding landmarks. Detecting unblocked components first helps us to focus on the informative area, resulting in a better performance. Additionally, to improve the performance of the above tasks, we design two models as follows: 1) we add AnchorAlign in the region proposal networks to accurately localize components and 2) we propose a two-branch model consisting classification branch and regression branch to detect landmark. Extensive evaluations on benchmark datasets indicate that our proposed approach is able to complete the multi-task facial detection and outperforms the state-of-the-art facial component and landmark detection algorithms.

Geometric Facial Components Feature Extraction for Facial Expression Recognition

Abstract: Facial expression recognition is an active research challenge in computer vision and artificial intelligence since facial expressions contribute non-verbal information in human communication. Capturing facial features become an important phase in facial recognition systems. Finding suitable feature descriptor is essential to determine the recognition results. We propose a novel geometric feature extraction method which apply simple calculation techniques for facial components to ensure the robustness for each variation of pose. Unlike any other features which require more efforts in a transformation process, the proposed method efficiently works directly on pixels basis. We apply our proposed features into a facial expression recognition system and validate emotion results on extended Cohn Kanade (CK+) emotion dataset and gives accuracy rate 93.67%.

Eulerian video magnification: a review

Abstract: Many important subtle changes in the environment are invisible to the naked human eyes. These subtle changes occur because of colour variations, such as blood flow in a human face that leads to face colour change, or motion variations, such as vena movement under human skin and vibration of buildings. The human eye requires optical microscopes to detect these variations. Alternatively, new technologies, such as high-speed imagery and computer processing, can be used to detect these variations. These computerised microscopes depend on computation rather than optical amplification to amplify subtle colour and motion changes in videos. The most popular technique to achieve computation-based microscope is the Eulerian video magnification (EVM). However, several challenges in EVM still need to be solved to meet the requirements of real time and video quality. This paper presents a comprehensive study of EVM methods and reviews the related literature. The strengths and drawbacks of existing works are discussed, and the important research fields and challenges in the area of EVM are concluded.

Detection of Prostate Cancer Using Radial/Axial Scanning of 2D Trans-rectal Ultrasound Images

Abstract: The search for improvement in the result of segmentation of regions of interest in medical images has continued to be a source of challenge to researchers Several research efforts have gone in to delineate regions of interest in the prostate gland from Trans-rectal ultrasound (TRUS) 2D-images In this work, we develop a fast algorithm based on radial/axial scanning of the pixels of the prostate gland image with the goal of detecting hyper-echoic pixels that are bound within the boundaries of the gland TRUS 2D-images The algorithm implements expert knowledge and utilizes the features extracted from the intensity of the TRUS images, primarily the relative intensity and gradient to delineate region of interest It employs radial/axial scanning of the image from common seed point automatically selected to detect the region of the gland and subsequently hyper-echoic pixels which indicate suspected cancerous tissue cites Evaluation of the algorithm performance was done by comparing detection result with that of expert radiologists The detection algorithm gave an average accuracy of 8855% and sensitivity of 7165%

An optimization of facial feature point detection program by using several types of convolutional neural network

Abstract: Detection of facial feature points is an important technique used for biometric authentication and facial expression estimation. A facial feature point is a local point indicating both ends of the eye, holes of the nose, and end points of the mouth in the face image. Many researches on face feature point detection have been done so far, but the accuracy of facial organ point detection is improving by the approach using Convolutional Neural Network (CNN). However, CNN not only takes time to learn but also the neural network becomes a complicated model, so it is necessary to improve learning time and detection accuracy. In this research, the improvement of the detection accuracy of the learning speed is improved by increasing the convolution layer.