The difficulty in precisely detecting and locating an ear within an image is the first step to tackle in an ear-based biometric recognition system, a challenge which increases in difficulty when working with variable photographic conditions. This is in part due to the irregular shapes of human ears, but also because of variable lighting conditions and the ever changing profile shape of an ear's projection when photographed. An ear detection system involving multiple convolutional neural networks and a detection grouping algorithm is proposed to identify the presence and location of an ear in a given input image. The proposed method matches the performance of other methods when analyzed against clean and purpose-shot photographs, reaching an accuracy of upwards of 98%, but clearly outperforms them with a rate of over 86% when the system is subjected to non-cooperative natural images where the subject appears in challenging orientations and photographic conditions.

No abstract

Texture classification poses a well known difficulty within computer vision systems. This paper reviews a method for image segmentation based on the classification of textures using artificial neural networks. The supervised machine learning system developed here is able to recognize and distinguish among multiple feature regions within one or more photographs, where areas of interest are characterized by the various patterns of color and shape they exhibit. The use of an enhancement filter to reduce sensitivity to illumination and orientation changes in images is explored, as well as various post-processing techniques to improve the classification results based on context grouping. Various applications of the system are examined, including the geographical segmentation of satellite images and a brief overview of the model's performance when employed on a real time video stream.

We present a technique for improving the speed of a convolutional neural network applied to large input images through the optimization of the sliding window approach. Meaningful performance gains and memory bandwidth reduction can be obtained by processing images in this manner, factors which play a crucial role in the deployment of deep neural networks within mobile devices.

The research offers a quite simple view of methods to classify edible and poisonous mushrooms. In fact, we are looking for not only classification methods but also for an application which supports experts' decisions. To achieve our aim, we will study different structures of neural nets and learning algorithms, and select the best one, according to the test results.