3D human pose and shape estimation is the foundation of analyzing human motion. However, estimating accurate and temporally consistent 3D human motion from a video remains a challenge. By now, most of the video-based methods for estimating 3D human pose and shape rely on unidirectional temporal features and lack more comprehensive information. To solve this problem, we propose a novel model "bidirectional temporal feature for human motion recovery" (BTMR), which consists of a human motion generator and a discriminator. The transformer-based generator effectively captures the forward and reverse temporal features to enhance the temporal correlation between frames and reduces the loss of spatial information. The motion discriminator based on Bi-LSTM can distinguish whether the generated pose sequences are consistent with the realistic sequences of the AMASS dataset. In the process of continuous generation and discrimination, the model can learn more realistic and accurate poses. We evaluate our BTMR on 3DPW and MPI-INF-3DHP datasets. Without the training set of 3DPW, BTMR outperforms VIBE by 2.4 mm and 14.9 mm/s 2 in PA-MPJPE and Accel metrics and outperforms TCMR by 1.7 mm in PA-MPJPE metric on 3DPW. The results demonstrate that our BTMR produces better accurate and temporal consistent 3D human motion.

It is essential for researchers to have a proper interpretation of remote sensing images (RSIs) and precise semantic labeling of their component parts. Although FCN(Fully Convolutional Networks)-like deep convolutional network architectures have been widely applied in the perception of autonomous cars, there are still two challenges in the semantic segmentation of RSIs. The first is to identify details in high-resolution images with complex scenes and to solve the class-mismatch issues; the second is to capture the edge of objects finely without being confused by the surroundings. HRNET has the characteristics of maintaining high-resolution representation by fusing feature information with parallel multi-resolution convolution branches. We adopt HRNET as a backbone and propose to incorporate the Class-Oriented Region Attention Module (CRAM) and Class-Oriented Context Fusion Module (CCFM) to analyze the relationships between classes and patch regions and between classes and local or global pixels, respectively. Thus, the perception capability of the model for the detailed part in the aerial image can be enhanced. We leverage these modules to develop an end-to-end semantic segmentation model for aerial images and validate it on the ISPRS Potsdam and Vaihingen datasets. The experimental results show that our model improves the baseline accuracy and outperforms some commonly used CNN architectures.

We propose a virtual crowd navigation approach based on deep reinforcement learning to improve the adaptability of virtual crowds in an unknown and complex environment. To address the problem of local optimum or slow iteration or even failure to converge due to sparse rewards in complex environments, we integrate the curiosity-driven mechanism, the key navigation points acquisition and the failure path penalty method in addition to combining long short-term memory networks, dynamic obstacle collision prediction with proximal policy optimization algorithms, which realizes the crowd navigation in a complex environment. The experimental results show that the proposed approach can simulate the motions of virtual crowds in various dynamic and complex scenarios without the environment modeling. The use of continuous action space also ensures that the movement trajectories of the virtual crowds are more realistic and natural. Furthermore, our approach can provide analysis and demonstration tools for a variety of independent collaborations such as competition and cooperation of group intelligence in an open and dynamic environment.