No abstract

Deep convolutional neural networks have achieved huge successes in application domains like object and face recognition. The performance gain is attributed to different facets of the network architecture such as: depth of the convolutional layers, activation function, pooling, batch normalization, forward and back propagation and many more. However, very little emphasis is made on the preprocessors. Unlike other recognition tasks, face recognition calls for careful consideration of the network's input because of image formation factors that can impact on the discriminative characteristics of the face. Therefore, in this paper, the network's preprocessing module is varied across different preprocessing approaches while keeping constant other facets of the deep network architecture, to investigate the contribution preprocessing makes to the network. Commonly used preprocessors are the data augmentation and normalization and are termed conventional preprocessors. Others are termed the unconventional preprocessors, they are: color space converters; HSV, CIE L*a*b* and YCBCR, grey-level resolution preprocessors; full-based and plane-based image quantization, illumination normalization and insensitive feature preprocessing using: histogram equalization (HE), local contrast normalization (LN) and complete face structural pattern (CFSP). To achieve fixed network parameters, CNNs with transfer learning is employed. The aim is to transfer knowledge from the high-level feature vectors of the Inception-V3 network (trained on ImageNet data) to offline preprocessed LFW target data; and features trained using the SoftMax classifier for face identification. The experiments show that the discriminative capability of the deep networks can be improved by preprocessing RGB data with HE, full-based and plane-based quantization, rgbGELog, and YCBCR, preprocessors before feeding it to CNNs. However, for best performance, the right setup of preprocessed data with augmentation and/or normalization is required. Overall, above 72% is achieved with plane-based image quantization which is found to increase the homogeneity of neighborhood pixels and utilizes a more reduced bit depth for better storage efficiency.INDEX TERMS Deep convolutional neural networks, face identification, preprocessing, transfer learning.

The aim of molecular communication (MC) is to develop communication systems for biochemical information interchange among nano devices. While there have been research efforts to model, design and analyze various MC network and devices, a lot still have to be done before we can develop and deploy such systems. Following the traditional physical layered architecture of communication networks, this paper presents design initiatives, open research issues and proffer suggestion towards modeling and designing diffusion-based MC devices.Specifically, we present biologically inspired modular theoretical design approach to modeling and designing MC devices. We also discussed various issues that should be considered in developing an accurate system-theoretic model of an MC system.

Abstract-Feature extraction task are primarily about making sense of the discriminative features/patterns of facial information and extracting them. However, most real world face images are almost always intertwined with imaging modality problems of which illumination is a strong factor. The compensation of the illumination factor using various illumination compensation techniques has been of interest in literatures with few emphasis on the adverse effect of the techniques to the task of extracting the actual discriminative features of a sample image for recognition. In this paper, comparative analyses of illumination compensation techniques for extraction of meaningful features for recognition using a single feature extraction method is presented. More also, enhancing red, green, blue gamma encoding (rgbGE) in the log domain so as to address the separability problem within a person class that most techniques incur is proposed. From experiments using plastic surgery sample faces, it is evident that the effect illumination compensation techniques have on face images after pre-processing is highly significant to recognition accuracy.