Optical Memory and Neural Networks

Automatic face recognition still suffers from some problems in the real-world scenarios such as occlusion. Hence, identifying the face from its partial appearance is a challenging issue as yet. To address this, issue many methods have been proposed using traditional feature extraction techniques. In this paper, a partial face recognition problem has been tackled through utilizing patch-wise matching with Convolutional Neural Network (CNN). Firstly, a gallery images are divided into local patches, and each patch is regarded as an independent image. Then, AlexNet architecture is utilized for training image patches. The Instance-To-Class (ITC) matching technique using K-Nearest Neighbour (KNN) algorithm specifies the class of the facial test image based on patch prediction. The notable contributions of our work are two-folds: the first one is employing ITC technique for patch prediction and the last one is adopting a deep learning technique for feature extraction and handling partial occlusion problem. The achieved accuracies on two de-facto datasets show that our method outperforms several existing methods that use hand-designed feature descriptors.

Face detection and recognition plays an important role in many occasions. This study explored the application of convolutional neural network in face detection and recognition. Firstly, convolutional neural network was briefly analyzed, and then a face detection model including three convolution layers, four pooling layers, introduction layers and three fully connected layers was designed. In face recognition, the self-learning convolutional neural network (CNN) model for global and local extended learning and Spatial Pyramid Pooling (SPP)-NET model were established. LFW data sets were used as model test samples. The results showed that the face detection model had an accuracy rate of 99%. In face recognition, the self-learning CNN model had an accuracy rate of 94.9% accuracy, and the SPP-Net model had an accuracy rate of 92.85%. It suggests that the face detection and recognition model based on convolutional neural network has good accuracy, and the face recognition efficiency of self-learning CNN model was better, which deserves further research and promotion.

For computer vision systems of autonomous vehicles, an important task is to ensure high reliability of visual information coming from on-board cameras. Frequent problems are contamination of the camera lens, its defocusing due to mechanical damage, image motion blur in low light conditions. In our work, we propose a novel neural network approach to the classification and unsupervised segmentation of visibility artifacts on monocular camera images. It is based on the compact classification deep neural network with an integrated modification of the gradient method for class activation map and segmentation mask generating. We present a new dataset named Visibility Artifacts containing over 22 300 images including six common artifacts: complete loss of camera visibility, strong or partial contamination, rain or snow drops, motion blur, defocus. To check the quality of artifact localization, a small test set with ground truth masks is additionally labeled. It allowed us to objectively quantitatively compare various methods for constructing class activation maps (CAMERAS, FullGrad, original and modified Grad-CAM, Layer-CAM), which demonstrated image segmentation quality above 54% mIoU without any supervision. This is a promising result. Experiments with the developed dataset demonstrated the superiority of the neural network classification method ResNet-18_U (with test accuracy of 99.37%), compared to more complex convolutional (ResNet-34, ResNeXt-50, EfficientNet-B0) and transformer (ViT-Ti, DeiT-Ti) neural networks. The code of the proposed method and the dataset are publicly available at https://github.com/vd-kuznetsov/CaUS_Visibility_Artifacts .

Characteristics of microdeformations caused by earthquakes and underground explosions were analyzed using the laser strainmeter measurements on the shore of the Japanese Sea. Packets of seismic waves from several earthquakes and from an artificial explosion were detected and analyzed with the help of Hilbert-Huang method for investigation of non-stationary and nonlinear processes. It was shown that the frequency (period) range of quasi harmonic seismic wave packets is very narrow and does not depend on their origin, but the length of the packets and of aftershocks is much shorter in the artificial explosion than in the natural ones. Physical interpretation is proposed for general and some specific features of the phenomena investigated.

this paper describes an algorithm for classification of roof materials using aerial photographs. Main advantages of the algorithm are proposed methods to improve prediction accuracy. Proposed methods includes: method of converting ImageNet weights of neural networks for using multi-channel images; special set of features of second level models that are used in addition to specific predictions of neural networks; special set of image augmentations that improve training accuracy. In addition, complete flow for solving this problem is proposed. The following content is available in open access: solution code, weight sets and architecture of the used neural networks. The proposed solution achieved second place in the competition “Open AI Caribbean Challenge”.

The films were drawn on the glass substrates from film-forming sols, created on the basis of chlorides or nitrates. These composites photocatalytic activity dependences on the films thickness and the packaging density of the sol disperse particles in a film layer have been established experimentally.

Modulation of white light emitted by the thermal source using simple optical systems capable to realize UWB information transfer are considered and realized. The experiments fulfilled permitted to estimate the minimum time resolution of the detection of signal with two-beams interferometer as 3.6 fs. For the effective time encoding method, the system based on Michelson echelon is proposed, its theory is developed using δ-wave approach for any angle of the beam incidence and scattering. Peculiarities of the echelon pulse response are analyzed. In addition, the simple method of digital encoding of the light flow with the pack of thin glasses of different thickness is demonstrated.

Bài viết trình bày mô hình thiết bị để đọc (hình dung) thông tin từ trường ẩn từ các hologram kết hợp với lớp từ quang. Các phương pháp hình thành hình ảnh từ trường trên tài liệu bảo mật và cách đọc chúng bằng các phương pháp từ quang được đề xuất. Đầu đọc sử dụng hiệu ứng Kerr cực từ từ quang cho phép quan sát trực quan "hiệu ứng nhấp nháy" từ hologram có lớp từ ẩn. Trong quá trình thực hiện, phân tích toán học về hiệu ứng Kerr hoặc hiệu ứng Faraday từ quang đã được tiến hành. Hình ảnh từ trường ẩn dựa trên: Ưu điểm của thiết bị: Sơ đồ quang của thiết bị bao gồm một nguồn sáng, bộ phân cực, bộ phân tích, hologram với các lớp từ quang, và nam châm cố định. Hologram được đặt giữa bộ phân cực và bộ phân tích.

The paper considers the use of convolutional neural networks for the concurrent recognition of the gender and age of a person by video records of his face. The emphasis is on the incorporation of the approach into mobile video analytics systems. We have investigated the fusion of decisions obtained during the processing of each video frame, including the use of the classifier committee based on Dempster-Shafer theory. We propose the novel age prediction method using the evaluation of the expectation of the most probable ages. We have compared existing neural-net models with a specially trained modification of the MobileNet convolution network with two outputs. The experimental results are given for such data collections as Kinect, IJB-A, Indian Movie and EmotiW. As compared with other conventional methods, our approach makes it possible to increase the age and gender recognition accuracy by 2–5% and 5–10% respectively.

We look into theoretical and experimental aspects of tribometric interaction between two dielectric substrates with identical surfaces in the course of rapid analysis of their cleanliness. It is shown that if the coefficient of sliding friction equals that of static friction the surface is can be defined as being of production grade. The tribometric interaction of the substrates is experimentally shown not to result in a mechanical damage of the surface under study. It is suggested that the device should be complemented by a diffraction grating that generates light pulses. Based on the grating parameters, such as light pulse duration, average sum of three pulses, and degree of their deviation from reference values, the level of substrate surface cleanliness is assessed. The use of the diffraction grating of period T = 63 μm and slit width b = 20 μm is shown to pro vide a 16-fold increase in the resolution of the tribometric device.