Gait planning of quadruped robots plays an important role in achieving less walking, including dynamic and static gait. In this article, a static and dynamic gait control method based on center of gravity stability ma...
详细信息
3D reconstruction from a single RGB image for urban scenes has been a foundation for safety-critical applications such as autonomous driving and city planning. It is essential to develop 3D reconstruction models that ...
详细信息
Layout estimation is one of fundamental tasks for understanding indoor scenes. In a single image of indoor scenes, key information, such as key points and boundaries for inferring the layout is often severely occluded...
详细信息
In view of the problem that the global optical flow algorithm cannot acquire accurate motion parameter estimation at a low-gradient value, an improved method has been presented in order to enhance the self-adaptive ab...
详细信息
It is well-known that the auxiliary information plays a key role in zero-shot classification. However, most of the existing popular methods do not make effective use of auxiliary information. To address this issue, we...
详细信息
ISBN:
(数字)9781728180281
ISBN:
(纸本)9781728180298
It is well-known that the auxiliary information plays a key role in zero-shot classification. However, most of the existing popular methods do not make effective use of auxiliary information. To address this issue, we propose an improved embedding model for zero-shot classification based on attention mechanism, called EMAM. In the proposed EMAM, we first add an attention mechanism to effectively extract the key information of auxiliary information in zero-shot classification. Then optimizes the objective function to improve the recognition rate of this model. Finally the experimental comparison is implemented on the standard zero-shot learning datasets. The experimental results demonstrate that our proposed EMAM not only verifies its validity, but also achieves good results.
This article focuses on the study of multi-resolution feature fusion methods for human pose estimation. Most of the existing pose estimation methods are based on high-resolution gradually reducing the resolution to le...
详细信息
ISBN:
(数字)9781728180281
ISBN:
(纸本)9781728180298
This article focuses on the study of multi-resolution feature fusion methods for human pose estimation. Most of the existing pose estimation methods are based on high-resolution gradually reducing the resolution to learn advanced semantic features, and then gradually recovering high-resolution features from the low-resolution semantic features to locate keypoints of the person instance. The process without fusion of rich features will inevitably lose more spatial and semantic information. The purpose of the multi-resolution feature fusion method proposed in this paper is to fuse spatial and semantic information on multiple scale feature maps, in addition, our network use the multi-receptive field fusion module on the same scale features to enhance feature extraction. Our methods can utilize more feature information to accurately locate keypoints.
As an executive component of robots, the precise control of electric cylinders is crucial for achieving accurate control of robots. This article proposes a position closed-loop electric cylinder control method based o...
详细信息
ISBN:
(数字)9798331506100
ISBN:
(纸本)9798331506117
As an executive component of robots, the precise control of electric cylinders is crucial for achieving accurate control of robots. This article proposes a position closed-loop electric cylinder control method based on three-orders active disturbance rejection control (ADRC). Firstly, model the electric cylinder system and determine its transfer function. Secondly, PID controllers and ADRC controllers were designed separately, and their stability was verified through the Routh criterion and Lyapunov method, respectively. Finally, the control effects of PID and ADRC were compared through three sets of experiments. The results show that compared to PID, ADRC is less affected by the increase in input signal frequency, has stronger anti-interference ability, and can effectively resist the influence of internal uncertainty.
We propose a scene text detection with feature aggregation and receptive field enhancement. It mainly improves the problem of wrong text segmentation. Specifically, a Feature Aggregation and Receptive Field Enhancemen...
详细信息
ISBN:
(数字)9781728180281
ISBN:
(纸本)9781728180298
We propose a scene text detection with feature aggregation and receptive field enhancement. It mainly improves the problem of wrong text segmentation. Specifically, a Feature Aggregation and Receptive Field Enhancement Module (FARE) adopts a ladder structure and constructs multi-scale feature aggregation with multiple branches to enhance the feature representation ability of texts of different scales. We add dilated convolution to expand the range of receptive field, which increases the detection of large-scale texts in low-level feature and obtains more accurate position information. Extensive experiments on the ICDAR 2015, ICDAR 2017 MLT, and CTW1500 show that the proposed method effectively improves the detection performance of the network. Notably, our proposed method has detected a precision of 87.4% on the curved text dataset CTW1500.
Semantic segmentation is one of the most important research directions in the field of computer vision, and has a wide range of applications for autonomous driving, medical imaging, intelligent security, etc. Unsuperv...
Semantic segmentation is one of the most important research directions in the field of computer vision, and has a wide range of applications for autonomous driving, medical imaging, intelligent security, etc. Unsupervised domain adaptation is the mainstream research topic in recent years, which can use a large number of labeled source samples to complete the segmentation task in target domain without labeled target samples. In this paper, we propose a prototype-guided unsupervised domain adaptation for semantic segmentation based on ProDA model. Due to lacking of labeled target samples and the prior probability, a prototype distance loss based on target domain is proposed to optimize the distribution of features by measuring the distance between features and the updated prototype and designing an adaptive threshold strategy. Meanwhile, a smoothing loss is proposed to alleviate the impact of source samples on our model and improve the prediction performance of the network. By conducting experiments on the GTA5 to Cityscapes scenarios, the results show that compared with the original model, the loss optimization improves mIoU by1.52.
Path planning and tracking algorithms are one of the cores of autonomous navigation technology for unmanned ground vehicle (UGV). In this paper, a local obstacle avoidance method based on event-triggered control is pr...
Path planning and tracking algorithms are one of the cores of autonomous navigation technology for unmanned ground vehicle (UGV). In this paper, a local obstacle avoidance method based on event-triggered control is proposed to ensure that the UGV avoids obstacles in the tracking process. First, the path planning and tracking control problem of the UGV is modeled and formulated. Second, the improvement and logical correlation of the relevant algorithms are described in detail. Finally, ajoint simulation of MATLAB and V-rep is performed. The simulation results show that event-triggered control can largely improve the reliability of the UGV path planning and tracking process, and verify the effectiveness of the algorithm.
暂无评论