scispace - formally typeset
Search or ask a question
Author

Kevin Roy

Bio: Kevin Roy is an academic researcher from Madras Institute of Technology. The author has contributed to research in topics: Facial recognition system & Computer science. The author has an hindex of 1, co-authored 1 publications receiving 3 citations.

Papers
More filters
Proceedings ArticleDOI
26 Mar 2015
TL;DR: Experimental procedures show that the proposed method facilitates recognition with mean accuracy of 85% even in case of partial occlusions, and the D-SIFT features (descriptors) are classified using Support Vector Machines (SVM).
Abstract: In this paper, a novel method for facial feature extraction and recognition using an optimized combination of Deformable Parts Model (DPM) and Dense Scale Invariant Feature Transform (D-SIFT) is proposed. Real time face recognition systems pose challenges such as the speed and responsiveness. When the basic SIFT algorithm is applied to the entire face, the number and location of the detected keypoints changes with illumination in real time. Moreover, occlusion results in the generation of unwanted keypoints which decreases accuracy. In general, more time is consumed for detection of keypoints. These challenges are addressed by Dense SIFT algorithm, as the number of keypoints and their locations are controlled. DPM allows reduction in dimensionality of features by considering eyes, nose and mouth patches as Regions of Interest (ROI). These patches can be independently recognized. The D-SIFT features (descriptors), extracted from the ROIs, are classified using Support Vector Machines (SVM). The proposed method is tested with the self-created and Caltech databases. Experimental procedures show that the proposed method facilitates recognition with mean accuracy of 85% even in case of partial occlusions.

3 citations

Proceedings ArticleDOI
31 Oct 2022
TL;DR: In this paper , a non convex optimization problem is formed with lasso regularization and solved via block coordinate descent (BCD) for joint signal estimation and topology identification with a nonlinear model, under the modelling assumption that signals are generated by a sparse VAR model in a latent space and then transformed by a set of invertible, componentwise nonlinearities.
Abstract: Topology identification from multiple time series has been proved to be useful for system identification, anomaly detection, denoising, and data completion. Vector autoregressive (VAR) methods have proved well in identifying directed topology from complex networks. The task of inferring topology in the presence of noise and missing observations has been studied for linear models. As a first approach to joint signal estimation and topology identification with a nonlinear model, this paper proposes a method to do so under the modelling assumption that signals are generated by a sparse VAR model in a latent space and then transformed by a set of invertible, component-wise nonlinearities. A non convex optimization problem is formed with lasso regularisation and solved via block coordinate descent (BCD). Initial experiments conducted on synthetic data sets show the identifying capability of the proposed method.
DOI
22 Dec 2022
TL;DR: In this paper , a nonlinear modeling technique for multiple time series that has a complexity similar to that of linear vector autoregressive (VAR), but it can account for nonlinear interactions for each sensor variable is proposed.
Abstract: Discovery of causal dependencies among time series has been tackled in the past either by using linear models, or using kernel- or deep learning-based nonlinear models, the latter ones entailing great complexity. This paper proposes a nonlinear modelling technique for multiple time series that has a complexity similar to that of linear vector autoregressive (VAR), but it can account for nonlinear interactions for each sensor variable. The modelling assumption is that the time series are generated in two steps: i) a VAR process in a latent space, and ii) a set of invertible nonlinear mappings applied component-wise, mapping each sensor variable into a latent space. Successful identification of the support of the VAR coefficients reveals the topology of the interconnected system. The proposed method enforces sparsity on the VAR coefficients and models the component-wise nonlinearities using invertible neural networks. To solve the estimation problem, a technique combining proximal gradient descent (PGD) and projected gradient descent is designed. Experiments conducted on real and synthetic data sets show that the proposed algorithm provides an improved identification of the support of the VAR coefficients, while improving also the prediction capabilities.

Cited by
More filters
Journal ArticleDOI
TL;DR: A multi-layer sparse coding model-based ship detection (MSCMSD) method, integrating bottom-up and top-down mechanisms, for ship detection with high-resolution remote-sensing images with higher accuracy than other state-of-the-art algorithms.
Abstract: Ship detection plays an important role in remote-sensing image processing. In this article, we propose a multi-layer sparse coding model-based ship detection (MSCMSD) method, integrating bottom-up and top-down mechanisms, for ship detection with high-resolution remote-sensing images. The multi-layer sparse coding model was designed to reveal the way how information is processed by human visual system. It is adopted in MSCMSD to detect candidate regions containing ships before any further processing. To detect ships from candidate regions, an omnidirectional solution is also proposed for deformable parts model-based ship detection. As demonstrated in the experiments, MSCMSD can detect ships from optical remote-sensing images with a higher accuracy than other state-of-the-art algorithms.

5 citations

Journal ArticleDOI
TL;DR: The results show thatFace Alignment, Face Detection, Feature Extraction, Feature Matching is superior to previously thought to be viable.
Abstract: Face recognition systems are in great demand for domestic and commercial applications. A novel feature extraction approach is proposed based on TanTrigg Lower Edge Directional Patterns for robust face recognition. Histogram of Orientated Gradients is used to detect faces and the facial landmarks are localized using Ensemble of Regression Trees. The detected face is rotated based on facial landmarks using affine transformation followed by cropping and resizing. TanTrigg preprocessor is used to convert the aligned face region into an illumination invariant region for better feature extraction. Eight directional Kirsch compass masks are convolved with the preprocessed face image. Feature descriptor is extracted by dividing the TTLEDP image into several sub-regions and concatenating the histograms of all the sub-regions. Chi-square distance metric is used to match faces from the trained feature space. The experimental results prove that the proposed TTLEDP feature descriptor has better recognition rate than existing methods, overcoming the challenges like varying illumination and noise KeywoRdS Face Alignment, Face Detection, Feature Extraction, Feature Matching

3 citations

Proceedings ArticleDOI
01 Aug 2017
TL;DR: A grid scheduling scheme which maintain two queues for GPUs and SW processors to obtain good scalability results and indicates that the parallel implementation on accelerators can achieve speedup up to 3x.
Abstract: Deformable part models (DPM) is a typical machine-learning based detection technique. It can achieve great success in detecting accuracy, but have compute-intensive tasks which severely restricts its utilization in many real world applications. In order to get high frame-rate for practical use, accelerators and grid computing infrastructure are needed. This paper propose a grid scheduling scheme which maintain two queues for GPUs and SW processors. The scheme obtain good scalability results through a lot of experiment. The experimental result on single processor node indicates that the parallel implementation on accelerators can achieve speedup up to 3x.