No abstract

Deep convolutional neural networks (CNNs) trained on visual objects have shown intriguing ability to predict some response properties of visual cortical neurons. However, the factors (e.g., if the model is trained or not, receptive field size) and computations (e.g., convolution, rectification, pooling, normalization) that give rise to such ability, at what level, and the role of intermediate processing stages in explaining changes that develop across areas of the cortical hierarchy are poorly understood. We focused on the sensitivity to textures as a paradigmatic example, since recent neurophysiology experiments provide rich data pointing to texture sensitivity in secondary (but not primary) visual cortex (V2). We initially explored the CNN without any fitting to the neural data and found that the first two layers of the CNN showed qualitative correspondence to the first two cortical areas in terms of texture sensitivity. We therefore developed a quantitative approach to select a population of CNN model neurons that best fits the brain neural recordings. We found that the CNN could develop compatibility to secondary cortex in the second layer following rectification and that this was improved following pooling but only mildly influenced by the local normalization operation. Higher layers of the CNN could further, though modestly, improve the compatibility with the V2 data. The compatibility was reduced when incorporating random rather than learned weights. Our results show that the CNN class of model is effective for capturing changes that develop across early areas of cortex, and has the potential to help identify the computations that give rise to hierarchical processing in the brain (code is available in GitHub).

No abstract

Abstract-The aim of this paper is to develop a real time vision-based facial expression recognition and adaptation system for human-computer interaction. Major objective of this research is to detect face, to identify and recognize user's facial expression using face image in real time and to be able to adapt with new user's facial expression. It also works on mixed race expression detection. It is based upon the eigenface algorithm. Which a small set of feature vectors are used to describe the variation between expression images. It is also being able to adapt new expression image in real time. The proposed system makes major contribution in implementing facial expression recognition and adaptation in real time. The facial expression recognition task is divided into two parts: first part consists of automatic face detection from video stream and preprocessing, second part consists of a classification step that employs Principal Component Analysis (PCA) to classify the expression into one of five categories. The algorithm has been tested using both static and dynamic images. The average precision and recall rate achieved by the system is about 88% for person specific recognition.Index Terms-Facial action coding system, facial expression recognition, hidden markov model (HMM), neural network (NN), principal component analysis (PCA).

2015).Offsetting obstacles of any shape for robot motion planning. Robotica, 33, pp 865-883 SUMMARYWe present an algorithm for offsetting the workspace obstacles of a circular robot. Our method has two major steps: It finds the raw offset curve for both lines and circular arcs, and then removes the global invalid loops to find the final offset. To generate the raw offset curve and remove global invalid loops, O(n) and O((n + k) log m) computational times are needed respectively, where n is the number of vertices in the original polygon, k is the number of self-intersections and m is the number of segments in the raw offset curve, where m ≤ n. Any local invalid loops are removed before generating the raw offset curve by invoking a pair-wise intersection detection test (PIDT). In the PIDT, two intersecting entities are checked immediately after they are computed, and if the test is positive, portions of the intersecting segments are removed. Our method works for conventional polygons as well as the polygons that contain circular arcs. Our algorithm is simple and very fast, as each sub-process of the algorithm can be completed in linear time except the last one, which is nearly linear. Therefore, the overall complexity of the algorithm is nearly linear. By applying our simple and efficient approach, offsetting obstacles of any shape make it possible to construct a configuration space that ensures optimized motion planning.