Occluded Pedestrian-Attribute Recognition for Video Sensors Using Group Sparsity.

Pedestrians are often obstructed by other objects or people in real-world vision sensors. These obstacles make pedestrian-attribute recognition (PAR) difficult; hence, occlusion processing for visual sensing is a key issue in PAR. To address this problem, we first formulate the identification of non-occluded frames as temporal attention based on the sparsity of a crowded video. In other words, a model for PAR is guided to prevent paying attention to the occluded frame. However, we deduced that this approach cannot include a correlation between attributes when occlusion occurs. For example, "boots" and "shoe color" cannot be recognized simultaneously when the foot is invisible. To address the uncorrelated attention issue, we propose a novel temporal-attention module based on group sparsity. Group sparsity is applied across attention weights in correlated attributes. Accordingly, physically-adjacent pedestrian attributes are grouped, and the attention weights of a group are forced to focus on the same frames. Experimental results indicate that the proposed method achieved 1.18% and 6.21% higher F1-scores than the advanced baseline method on the occlusion samples in DukeMTMC-VideoReID and MARS video-based PAR datasets, respectively.

Significance of physical factors on activities of daily living in patients with tetraplegia after spinal cord injury: a retrospective study.

BACKGROUND: Tetraplegia is a debilitating sequela of spinal cord injury (SCI). However, comprehensive approaches for determining the influence of various factors on activities of daily living (ADL) in patients with tetraplegia are limited. Therefore, this study aimed to determine the influence of physical factors on ADL in patients with tetraplegia after adjusting for demographic, SCI-related, and cognitive factors. METHODS: This retrospective cross-sectional study enrolled 201 patients with tetraplegia who underwent inpatient rehabilitation at the National Rehabilitation Center in South Korea between 2019 and 2021. Patients' mean age was 50.5 years (standard deviation, 16.3), and 170 (84.6%) were men. The Korean Spinal Cord Independence Measure III (K-SCIM III) was used as the main outcome measure to assess patients' ADL ability. Hierarchical multiple regression modeling was conducted with K-SCIM as the dependent variable to examine the level of functioning and relative influencing factors. RESULTS: Upper-extremity motor score (UEMS), upper-extremity spasticity and sitting balance scores were significant predictors of self-care; lower-extremity motor score (LEMS), musculoskeletal pain of shoulder, and sitting balance were significant predictors of respiratory and sphincter management; UEMS, LEMS, and sitting balance score were significant predictors of mobility; and UEMS, LEMS, musculoskeletal pain of shoulder, and sitting balance scores were significant predictors of the K-SCIM III total score after adjustment for demographic, SCI-related, and cognitive factors. CONCLUSIONS: Physical factors had the greatest impact on all subscores and the K-SCIM III total score. Upper- and lower-extremity muscle strength and sitting balance significantly affected functional ability across all subscores.

Action-Driven Visual Object Tracking With Deep Reinforcement Learning.

In this paper, we propose an efficient visual tracker, which directly captures a bounding box containing the target object in a video by means of sequential actions learned using deep neural networks. The proposed deep neural network to control tracking actions is pretrained using various training video sequences and fine-tuned during actual tracking for online adaptation to a change of target and background. The pretraining is done by utilizing deep reinforcement learning (RL) as well as supervised learning. The use of RL enables even partially labeled data to be successfully utilized for semisupervised learning. Through the evaluation of the object tracking benchmark data set, the proposed tracker is validated to achieve a competitive performance at three times the speed of existing deep network-based trackers. The fast version of the proposed method, which operates in real time on graphics processing unit, outperforms the state-of-the-art real-time trackers with an accuracy improvement of more than 8%.