Research Output
Cloud-based embedded system for object detection and recognition
Face Recognition System Based on Fusion Features of Local Methods Using CCA
A review of optimization algorithms in SVM parameters
1 - 3 of 3
No Thumbnail Available
Publication
Object detection and recognition techniques require large image datasets, memory, a workstation with specific graphics processing capability to train the algorithm and might have high power consumption. Embedded platforms on the other hand are characterized by portability, low power consumption and space, and energy resources making the deployment of such algorithms on them difficult. In order to overcome these drawbacks, cloud-based processing embedded system for object detection and recognition is proposed in this work. The system consists of an image acquisition device set up using embedded board and camera to capture, process and send images to the remote computer via cloud storage platform. This cloud platform serves as an interface between the embedded board and the remote computer. The detection algorithm of Faster R-CNN is executed on the remote computer and is trained and validated with 3000 images obtained from ImageNet. The training of the algorithm aims to detect five classes of object. The proposed system was validated off-line and have achieved a mean Average Precision (mAP) of 0.67. The performance of entire system procedure took about 45 seconds and have obtained an average confidence score of 0.86.
Publication
Information fusion is a solution espoused for enhancing a pattern recognition system's performance. A single representation précises the information and presents a single cue on the data; thus, information fusion is said to be more prolific as every feature set depicts a different outlook on the actual dataset. This paper recommends a face recognition system by utilizing fusion features of two local descriptor approaches. Firstly, blending of two most effective local face features, namely Gabor transform features and Local Binary Pattern (LBP), renders significantly improved performance compared to either individually: they complement each other wherein small appearance details are captured by LBP, while Gabor includes encoding facial shape for a wider range of scales. Secondly, to the combined feature vector, applying of the Canonical Correlation Analysis method (CCA) is done in order to extract discriminant characteristics for recognition. Lastly, a support vector machine (SVM) is deployed for classification, and K-nearest neighbor (K-NN) is utilized for feature matching. The technique is assessed against many challenging face datasets such as Yale database, with encouraging outcomes.
No Thumbnail Available
Publication
The SVM is a widely known machine learning, which is very useful for regression applications and pattern classification. These machines have been used successfully in several domains to address numerous real-world challenges. In this context, parameter optimisation for an SVM is a widely researched topic, which has attracted attention from several research domains. Algorithms facilitating optimisation have been of greater interest compared to other algorithms. Algorithmic approaches allow the optimal parameters for an SVM to be determined, after which the model can be adapted for several other applications. During the last two decades, several enhancements have been brought about to facilitate better optimisation of SVM models to offer enhanced performance. This paper focuses on the several algorithms currently employed to optimise support vector machines in their basic and modified forms. This paper comprises a comprehensive analysis of algorithms and aims to ascertain the present challenges relating to algorithms used for SVM parameter optimisation. This study cannot evaluate all the details; however, the significant theoretical aspects are covered using references to existing literature.