Object detection is one of the core tasks of computer vision, and bounding box (bbox) regression is one of the basic tasks of object detection. In recent years of related research, bbox regression is often used in the Intersection over Union (IoU) loss and its improved version. In this paper, for the first time, we introduce the Dice coefficient into the regression loss calculation and propose a new measure which is superior to and can replace the IoU. We define three properties of the new measure and prove the theory by mathematical reasoning and analysis of the existing work. This paper also proposes the N-IoU regression loss family. And the superiority of the N-IoU regression loss family is proved by designing simulation experiments and comparative experiments. The main results of this paper are: (1) The proposed new measure is better than IoU which can be used to evaluate bounding box regression, and the three properties of the new measure can be used as a broad criterion for the design of regression loss functions; and (2) we propose N-IoU loss. The parameter n of N-IOU can be debugged, which can be widely adapted to different application scenarios with higher flexibility, and the regression performance is better.

Vision transformers (ViTs) have become one of the dominant frameworks for vision tasks in recent years because of their ability to efficiently capture long-range dependencies in image recognition tasks using self-attention. In fact, both CNNs and ViTs have advantages and disadvantages in vision tasks, and some studies suggest that the use of both may be an effective way to balance performance and computational cost. In this paper, we propose a new hybrid network based on CNN and transformer, using CNN to extract local features and transformer to capture long-distance dependencies. We also proposed a new feature map resolution reduction based on Discrete Cosine Transform and self-attention, named DCT-Attention Down-sample (DAD). Our DctViT-L achieves 84.8% top-1 accuracy on ImageNet 1K, far outperforming CMT, Next-ViT, SpectFormer and other state-of-the-art models, with lower computational costs. Using DctViT-B as the backbone, RetinaNet can achieve 46.8% mAP on COCO val2017, which improves mAP by 2.5% and 1.1% with less calculation cost compared with CMT-S and SpectFormer as the backbone.

This paper describes a pedestrian re-identification algorithm, which was developed by integrating semi-supervised learning and similarity-preserving generative adversarial networks (SPGAN). The pedestrian re-identification task aimed to rapidly capture the same target using different cameras. Importantly, this process can be applied in the field of security. Because real-life environments are complex, the number of detected identities is uncertain, and the cost of manual labeling is high; therefore, it is difficult to apply the re-identification model based on supervised learning in real-life scenarios. To use the existing labeled dataset and a large amount of unlabeled data in the application environment, this report proposes a semi-supervised pedestrian re-identification model, which combines a teacher–student model with SPGAN. SPGAN was used to reduce the difference between the target domain and the source domain by transferring the style of the labeled dataset from the source domain. Additionally, the dataset from the source domain was used after the style transfer to pre-train the model; this enabled the model to adapt more rapidly to the target domain. The teacher–student model and the transformer model were then employed to generate soft pseudo-labels and hard pseudo-labels (via iterative training) and to update the parameters through distillation learning. Thus, it retained the learned features while adapting to the target domain. Experimental results indicated that the maps of the applied method on the Market-to-Duke, Duke-to-Market, Market-to-MSMT, and Duke-to-MSMT domains were 70.2, 79.3, 30.2, and 33.4, respectively.

This paper presents a method for measuring aircraft landing gear angles based on a monocular camera and the CAD aircraft model. Condition monitoring of the aircraft landing gear is a prerequisite for the safe landing of the aircraft. Traditional manual observation has an intense subjectivity. In recent years, target detection models dependent on deep learning and pose estimation methods relying on a single RGB image have made significant progress. Based on these advanced algorithms, this paper proposes a method for measuring the actual angles of landing gears in two-dimensional images. A single RGB image of an aircraft is inputted to the target detection module to obtain the key points of landing gears. The vector field network votes the key points of the fuselage after extraction and scale normalization of the pixels inside the aircraft prediction box. Knowing the pixel position of the key points and the constraints on the aircraft, the angle between the landing gear and fuselage plane can be calculated even without depth information. The vector field loss function is improved based on the distance between pixels and key points, and synthetic datasets of aircraft with different angle landing gears are created to verify the validity of the proposed algorithm. The experimental results show that the mean error of the proposed algorithm for the landing gears is less than 5 degrees on the light-varying dataset.

This paper proposes an action recognition algorithm based on the capsule network and Kalman filter called “Reading Pictures Instead of Looking” (RPIL). This method resolves the convolutional neural network’s over sensitivity to rotation and scaling and increases the interpretability of the model as per the spatial coordinates in graphics. The capsule network is first used to obtain the components of the target human body. The detected parts and their attribute parameters (e.g., spatial coordinates, color) are then analyzed by Bert. A Kalman filter analyzes the predicted capsules and filters out any misinformation to prevent the action recognition results from being affected by incorrectly predicted capsules. The parameters between neuron layers are evaluated, then the structure is pruned into a dendritic network to enhance the computational efficiency of the algorithm. This minimizes the dependence of in-depth learning on the random features extracted by the CNN without sacrificing the model’s accuracy. The association between hidden layers of the neural network is also explained. With a 90% observation rate, the OAD dataset test precision is 83.3%, the ChaLearn Gesture dataset test precision is 72.2%, and the G3D dataset test precision is 86.5%. The RPILNet also satisfies real-time operation requirements (>30 fps).