Enhancing the environmental perception of autonomous vehicles (AVs) in intelligent transportation systems requires computer vision technology to be effective in detecting objects and obstacles, particularly in adverse weather conditions. Adverse weather circumstances present serious difficulties for object-detecting systems, which are essential to contemporary safety procedures, infrastructure for monitoring, and intelligent transportation. AVs primarily depend on image processing algorithms that utilize a wide range of onboard visual sensors for guidance and decisionmaking. Ensuring the consistent identification of critical elements such as vehicles, pedestrians, and road lanes, even in adverse weather, is a paramount objective. This paper not only provides a comprehensive review of the literature on object detection (OD) under adverse weather conditions but also delves into the ever-evolving realm of the architecture of AVs, challenges for automated vehicles in adverse weather, the basic structure of OD, and explores the landscape of traditional and deep learning (DL) approaches for OD within the realm of AVs. These approaches are essential for advancing the capabilities of AVs in recognizing and responding to objects in their surroundings. This paper further investigates previous research that has employed both traditional and DL methodologies for the detection of vehicles, pedestrians, and road lanes, effectively linking these approaches with the evolving field of AVs. Moreover, this paper offers an in-depth analysis of the datasets commonly employed in AV research, with a specific focus on the detection of key elements in various environmental conditions, and then summarizes the evaluation matrix. We expect that this review paper will help scholars to gain a better understanding of this area of research.

In smart cities, target detection is one of the major issues in order to avoid traffic congestion. It is also one of the key topics for military, traffic, civilian, sports, and numerous other applications. In daily life, target detection is one of the challenging and serious tasks in traffic congestion due to various factors such as background motion, small recipient size, unclear object characteristics, and drastic occlusion. For target examination, unmanned aerial vehicles (UAVs) are becoming an engaging solution due to their mobility, low cost, wide field of view, accessibility of trained manipulators, a low threat to people’s lives, and ease to use. Because of these benefits along with good tracking effectiveness and resolution, UAVs have received much attention in transportation technology for tracking and analyzing targets. However, objects in UAV images are usually small, so after a neural estimation, a large quantity of detailed knowledge about the objects may be missed, which results in a deficient performance of actual recognition models. To tackle these issues, many deep learning (DL)-based approaches have been proposed. In this review paper, we study an end-to-end target detection paradigm based on different DL approaches, which includes one-stage and two-stage detectors from UAV images to observe the target in traffic congestion under complex circumstances. Moreover, we also analyze the evaluation work to enhance the accuracy, reduce the computational cost, and optimize the design. Furthermore, we also provided the comparison and differences of various technologies for target detection followed by future research trends.

In the ever-expanding online fashion market, businesses in the clothing sales sector are presented with substantial growth opportunities. To utilize this potential, it is crucial to implement effective methods for accurately identifying clothing items. This entails a deep understanding of customer preferences, niche markets, tailored sales strategies, and an improved user experience. Artificial intelligence (AI) systems that can recognize and categorize clothing items play a crucial role in achieving these objectives, empowering businesses to boost sales and gain valuable customer insights. However, the challenge lies in accurately classifying diverse attire items in a rapidly evolving fashion landscape. Variations in styles, colors, and patterns make it difficult to consistently categorize clothing. Additionally, the quality of images provided by users varies widely, and background clutter can further complicate the task of accurate classification. Existing systems may struggle to provide the level of accuracy needed to meet customer expectations. To address these challenges, a meticulous dataset preparation process is essential. This includes careful data organization, the application of background removal techniques such as the GrabCut Algorithm, and resizing images for uniformity. The proposed solution involves a hybrid approach, combining the strengths of the ResNet152 and EfficientNetB7 architectures. This fusion of techniques aims to create a classification system capable of reliably distinguishing between various clothing items. The key innovation in this study is the development of a Two-Objective Learning model that leverages the capabilities of both ResNet152 and EfficientNetB7 architectures. This fusion approach enhances the accuracy of clothing item classification. The meticulously prepared dataset serves as the foundation for this model, ensuring that it can handle diverse clothing items effectively. The proposed methodology promises a novel approach to image identification and feature extraction, leading to impressive classification accuracy of 94%, coupled with stability and robustness.

Due to the visual properties of reflection and refraction, RGB-D cameras cannot accurately capture the depth of transparent objects, leading to incomplete depth maps. To fill in the missing points, recent studies tend to explore new visual features and design complex networks to reconstruct the depth, however, these approaches tremendously increase computation, and the correlation of different visual features remains a problem. To this end, we propose an efficient depth completion network named DistillGrasp which distillates knowledge from the teacher branch to the student branch. Specifically, in the teacher branch, we design a position correlation block (PCB) that leverages RGB images as the query and key to search for the corresponding values, guiding the model to establish correct correspondence between two features and transfer it to the transparent areas. For the student branch, we propose a consistent feature correlation module (CFCM) that retains the reliable regions of RGB images and depth maps respectively according to the consistency and adopts a CNN to capture the pairwise relationship for depth completion. To avoid the student branch only learning regional features from the teacher branch, we devise a distillation loss that not only considers the distance loss but also the object structure and edge information. Extensive experiments conducted on the ClearGrasp dataset manifest that our teacher network outperforms state-of-the-art methods in terms of accuracy and generalization, and the student network achieves competitive results with a higher speed of 48 FPS. In addition, the significant improvement in a real-world robotic grasping system illustrates the effectiveness and robustness of our proposed system.

Self-supervised monocular depth estimation has achieved notable success under daytime conditions. However, its performance deteriorates markedly at night due to low visibility and varying illumination, e.g., insufficient light causes textureless areas, and moving objects bring blurry regions. To this end, we propose a self-supervised framework named DASP that leverages spatiotemporal priors for nighttime depth estimation. Specifically, DASP consists of an adversarial branch for extracting spatiotemporal priors and a self-supervised branch for learning. In the adversarial branch, we first design an adversarial network where the discriminator is composed of four devised spatiotemporal priors learning blocks (SPLB) to exploit the daytime priors. In particular, the SPLB contains a spatial-based temporal learning module (STLM) that uses orthogonal differencing to extract motion-related variations along the time axis and an axial spatial learning module (ASLM) that adopts local asymmetric convolutions with global axial attention to capture the multiscale structural information. By combining STLM and ASLM, our model can acquire sufficient spatiotemporal features to restore textureless areas and estimate the blurry regions caused by dynamic objects. In the self-supervised branch, we propose a 3D consistency projection loss to bilaterally project the target frame and source frame into a shared 3D space, and calculate the 3D discrepancy between the two projected frames as a loss to optimize the 3D structural consistency and daytime priors. Extensive experiments on the Oxford RobotCar and nuScenes datasets demonstrate that our approach achieves state-of-the-art performance for nighttime depth estimation. Ablation studies further validate the effectiveness of each component.