Continuous Viewpoint Planning in Conjunction with Dynamic Exploration for Active Object Recognition

Active object recognition (AOR) aims at collecting additional information to improve recognition performance by purposefully adjusting the viewpoint of an agent. How to determine the next best viewpoint of the agent, i.e., viewpoint planning (VP), is a research focus. Most existing VP methods perform viewpoint exploration in the discrete viewpoint space, which have to sample viewpoint space and may bring in significant quantization error. To address this challenge, a continuous VP approach for AOR based on reinforcement learning is proposed. Specifically, we use two separate neural networks to model the VP policy as a parameterized Gaussian distribution and resort the proximal policy optimization framework to learn the policy. Furthermore, an adaptive entropy regularization based dynamic exploration scheme is presented to automatically adjust the viewpoint exploration ability in the learning process. To the end, experimental results on the public dataset GERMS well demonstrate the superiority of our proposed VP method.

COVID-19 Social Distance Violation and Face Mask Detection in Workplaces

Social distancing is a suggested arrangement by the World Health Organization (WHO) to limit the spread of COVID-19 in broad daylight places. Most of governments and public wellbeing specialists have set the 2-meter physical removing as a compulsory security measure in retail outlets, schools, and other covered regions. In this exploration, we foster a conventional Deep Neural Network-Based model for mechanized individuals’ identification, following, and between individuals’ distances assessment in the group, utilizing basic CCTV surveillance cameras. The proposed model incorporates a YOLOv4-based system and opposite viewpoint planning for exact individuals’ identification and social removing checking in testing conditions, including individual’s impediment, incomplete perceivability, and lighting varieties. We additionally give an online danger appraisal conspire by factual examination of the Spatio-transient information from the moving directions and the pace of social removing infringement. We distinguish high-hazard zones with the most noteworthy chance of infection spread and diseases. This may assist specialists with updating the design of a public spot or to play it safe activities to relieve high-hazard zones. The effectiveness of the proposed approach is assessed on the Oxford Town Center dataset, with prevalent execution as far as precision and speed contrasted with three bests in class techniques.