Semi-supervised approaches for crowd counting attract attention, as the fully supervised paradigm is expensive and laborious due to its request for a large number of images of dense crowd scenarios and their annotations. This paper proposes a spatial uncertainty-aware semi-supervised approach via regularized surrogate task (binary segmentation) for crowd counting problems. Different from existing semisupervised learning-based crowd counting methods, to exploit the unlabeled data, our proposed spatial uncertaintyaware teacher-student framework focuses on high confident regions' information while addressing the noisy supervision from the unlabeled data in an end-to-end manner. Specifically, we estimate the spatial uncertainty maps from the teacher model's surrogate task to guide the feature learning of the main task (density regression) and the surrogate task of the student model at the same time. Besides, we introduce a simple yet effective differential transformation layer to enforce the inherent spatial consistency regularization between the main task and the surrogate task in the student model, which helps the surrogate task to yield more reliable predictions and generates high-quality uncertainty maps. Thus, our model can also address the task-level perturbation problems that occur spatial inconsistency between the primary and surrogate tasks in the student model. Experimental results on four challenging crowd counting datasets demonstrate that our method achieves superior performance to the state-of-the-art semi-supervised methods. Code is available at : https://github.com/ smallmax00/SUA_crowd_counting

Background COVID-19 has spread worldwide and generated tremendous stress on human beings. Unfortunately, it is often hard for distressed individuals to access mental health services under conditions of restricted movement or even lockdown. Objective The study first aims to develop an online digital intervention package based on a commercially released coloring game. The second aim is to test the effectiveness of difference intervention packages for players to increase subjective well-being (SWB) and reduce anxiety during the pandemic. Methods An evidence-based coloring intervention package was developed and uploaded to an online coloring game covering almost 1.5 million players worldwide in January 2021. Players worldwide participated to color either 4 rounds of images characterized by awe, pink, nature, and blue or 4 rounds of irrelevant images. Participants' SWB and anxiety and the perceived effectiveness of the game in reducing anxiety (subjective effectiveness [SE]) were assessed 1 week before the intervention (T1), after the participants completed pictures in each round (T2-T5), and after the intervention (T6). Independent 2-tailed t tests were conducted to examine the general intervention (GI) effect and the intervention effect of each round. Univariate analysis was used to examine whether these outcome variables were influenced by the number of rounds completed. Results In total, 1390 players worldwide responded and completed at least 1 assessment. Overall, the GI group showed a statistical significantly greater increase in SWB than the general control (GC) group (N=164, t162=3.59, Cohen d=0.59, 95% CI 0.36-1.24, P<.001). Compared to the control group, the best effectiveness of the intervention group was seen in the awe round, in which the increase in SWB was significant (N=171, t169=2.51, Cohen d=0.39, 95% CI 0.10-0.82, P=.01), and players who colored all 4 pictures had nearly significant improvements in SWB (N=171, F4,170=2.34, partial ŋ2=0.053, P=.06) and a significant decrease in anxiety (N=171, F4,170=3.39, partial ŋ2=0.075, P=.01). Conclusions These data indicate the effectiveness of online psychological interventions, such as coloring games, for mental health in the specific period. They also show the feasibility of applying existing commercial games embedded with scientific psychological interventions that can fill the gap in mental crises and services for a wider group of people during the pandemic. The results would inspire innovations to prevent the psychological problems caused by public emergencies and encourage more games, especially the most popular ones, to take more positive action for the common crises of humankind.

Blended learning is an information-based teaching method. This method is applied in network marketing course and verifying its teaching effect is a problem worthy of research. By combining blended learning-related theories, student features, and course requirements, a WeChat-based blended learning pattern applicable to network marketing course was proposed in this study. A teaching design of WeChat-based blended learning was created, followed by a teaching experiment. Quantitative analysis was used to verify the teaching results of the experimental and control classes. Results demonstrate that the learning effect of experimental class adopting WeChat-based blended learning is superior to that of the control class adopting traditional teaching method. WeChat-based blended learning lengthens extracurricular learning time, enhances students’ learning interest, and increases the opportunity of student–teacher exchange. This study provides references for developing WeChat-based hybrid learning.

Segmentation is an essential operation of image processing. The convolution operation suffers from a limited receptive field, while global modelling is fundamental to segmentation tasks. In this paper, we apply graph convolution into the segmentation task and propose an improved Laplacian. Different from existing methods, our Laplacian is data-dependent, and we introduce two attention diagonal matrices to learn a better vertex relationship. In addition, it takes advantage of both region and boundary information when performing graph-based information propagation. Specifically, we model and reason about the boundary-aware region-wise correlations of different classes through learning graph representations, which is capable of manipulating long range semantic reasoning across various regions with the spatial enhancement along the object's boundary. Our model is well-suited to obtain global semantic region information while also accommodates local spatial boundary characteristics simultaneously. Experiments on two types of challenging datasets demonstrate that our method outperforms the state-ofthe-art approaches on the segmentation of polyps in colonoscopy images and of the optic disc and optic cup in colour fundus images.

No abstract