Zhengyou Wang

Object detection, as a fundamental task, forms the cornerstone of intelligent applications in both UAV surveillance and satellite remote sensing. While most prior works concentrate on solving object scale and rotation angle variance caused by altitude changes, the spatial misalignment stemming from the differing demands of classification subtask and regression subtask also plays a critical role. To tackle these problems, a novel deep-guided dual-task collaborative learning framework is proposed. This framework integrates two key modules: deep-guided collaborative feature fusion (DGC-FF) and dual-task collaborative feature alignment (DTC-FA). DGC-FF effectively integrates fine-grained spatial and semantic information to enhance the network’s multi-scale perception capability. DTC-FA alleviates spatial misalignment between classification and regression branches through collaborative feature alignment and incorporates a rotation-aware detection branch to adapt to varying object orientations. Experimental results show that the proposed method achieves mAP@0.5 of 79.3% on the DroneVehicle dataset and mAP@0.5 of 81.6% on the DIOR-R dataset. The proposed method not only outperforms all compared methods in accuracy but also strikes a favorable efficiency–accuracy balance with an inference rate of 55–58 FPS.

Multi-object tracking (MOT) is an important problem in computer vision that has a wide range of applications. Currently, object occlusion detecting is still a serious challenge in multi-object tracking tasks. In this paper, we propose a method to simultaneously improve occluded object detection and occluded object tracking, as well as propose a tracking method for when the object is completely occluded. First, motion track prediction is utilized to improve the upper limit of occluded object detection. Then, the spatio-temporal feature information between the object and the surrounding environment is used for multi-object tracking. Finally, we use the hypothesis frame to continuously track the completely occluded object. Our study shows that we achieve competitive performances compared to the current state-of-the-art methods on popular multi-object tracking benchmarks such as MOT16, MOT17, and MOT20.

Applying convolutional neural networks (CNNs) to triangular meshes has always been a challenging task. Because of the complex structure of the meshes, most of the existing methods apply CNNs indirectly to them, and require complex preprocessing or transformation of the meshes. In this paper, we propose a novel face-based CNN, which can be directly applied to triangular meshes with arbitrary connectivity by defining face convolution and pooling. The proposed approach takes each face of the meshes as the basic element, similar to CNNs with pixels of 2D images. First, the intrinsic features of the faces are used as the input features of the network. Second, a sort convolution operation with adjustable convolution kernel sizes is constructed to extract the face features. Third, we design an approximately uniform pooling operation by learnable face collapse, which can be applied to the meshes with arbitrary connectivity, and we directly use its inverse operation as unpooling. Extensive experiments show that the proposed approach is comparable to, or can even outperform, state-of-the-art methods in mesh classification and mesh segmentation.

Face age estimation has been widely used in video surveillance, human-computer interaction, market analysis, image processing analysis, and many fields. There are several problems that need to be solved in image-based face age estimation: (1) redundant information of age characteristics; (2) limitations of age estimation methods in solving age estimation problems; (3) the performance of age estimation models being also affected by gender factors. This paper proposes CA-XTree network. Firstly, features are extracted through the convolution layer and then combined with the local channel attention module to strengthen the ability of age feature information interaction between different channels. Secondly, extracted features are inputted into the recommendation score function to obtain the recommendation score, by combining the recommendation score with the gradient ascending regression tree. The lifting tree processed loss function is the mean square loss function, and the final age value is obtained by the leaf node. This paper improves state of the art for image classification on MORPH and CACD datasets. The advantage of our model is that it is easy to implement and has no excess memory overhead. In the age dataset CACD, the mean absolute error (MAE) has reached 4.535 and cumulative score (CS) has reached 63.53, respectively.

Occlusion pedestrian detection is an important and difficult task in pedestrian detection. At present, the main method to deal with occlusion pedestrian detection usually adopts pedestrian parts or human body relationship methods. However, in the scene of crowd occlusion or severe pedestrian occlusion, only small parts of the body can be used for detection. Pedestrian parts or human body relationship methods cannot effectively address these issues. In view of the above problems, this paper abandoned the occlusion processing method of pedestrian parts or human body relationship. Considering that it is difficult to establish the relationship between parts and key points. The scale of visible parts of the occlusion pedestrian is small, and the scale of no occlusion pedestrian and occlusion pedestrian in the same picture is different. A multiscale feature attention fusion network named parallel feature fusion with CBAM (PFF-CB) is proposed for occlusion pedestrian detection. Feature information of different scales can be integrated effectively in the PFF-CB module. PFF-CB module uses a convolutional block attention module (CBAM) to enhance the important feature information in space and channel. A parallel feature fusion module based on FPN is used to enhance key features. The performance of the proposed module was tested on two common data sets of occlusion pedestrians with different occlusion types. The results show that the PFF-CB module makes a good performance in occlusion pedestrian detection tasks.