Classification of traffic signs with partial occlusions is important for traffic sign maintenance and inventory systems. It is also important to help drivers identify possible traffic signs in time. Motivated by human cognitive processes in identifying an occluded sign, a novel structure is designed to explicitly handle occluded samples in this paper. Occlusion maps are analyzed for possible occluded signs, and a new occlusion descriptor is proposed to distinguish occluded signs from negative samples. A series of tests shows that the developed method could effectively handle samples with partial occlusions and thus reduce the missed detections caused by occlusions. The developed method could also be easily used for any other object detection.

Reliable estimation of people in public areas is an important problem in visual surveillance. Although there is a lot of research on people counting in recent years, most of them consider a small crowd of people without many serious occlusions. Some of them have a lot of particular requirements, like people are moving, the background is smooth or the image resolution is high. This paper aims to estimate the number of people in a complicated scenario, which has around one hundred persons in an outdoors event. Several people counting methods based on crowd density are considered to find the relationship between the foreground pixels and the number of people in the large crowd. The best estimation result is from the method that considers two types of foreground pixels: those that come from relatively stationary crowd, and those that come from moving people. In an evaluation of three developed methods over 51 cases, the best average error is around 10%. All the proposed methods do not have any special requirements on the resolution of the input video.

In this paper, our aim is to segment a foreground region into individual persons in crowded scenes. We will focus on the combination of multiple clues for crowd segmentation. To ensure a wide range of applications, few assumptions are needed on the scenarios. In the developed method, crowd segmentation is formulated as a process to group the feature points with a human model. It is assumed that a foreground region has been detected and that an informative foreground contour is not required. The approach adopts a block-based implicit shape model (B-ISM) to collect some typical patches from a human being and assess the possibility of their occurrence in each part of a body. The combination of appearance cues with coherent motion of the feature points in each individual is considered. Some results based on the USC-Campus sequence and the CAVIAR data set have been shown. The contributions of this paper are threefold. First, a new B-ISM model is developed, and it is combined with joint occlusion analysis for crowd segmentation. The requirement for an accurate foreground contour is reduced. In addition, ambiguity in a dense area can be handled by collecting the evidences inside the crowd region based on the B-ISM. Furthermore, motion cues-which are coherent moving trajectories of feature points from individuals"are combined with appearance cues to help segment the foreground region into individuals. The usage of motion cues can be an effective supplement to appearance cues, particularly when the background is cluttered or the crowd is dense. Third, three features have been proposed to distinguish points on rigid body parts from those with articulated movements. Coherent motion of feature points on each individual can be more reliably identified by excluding points with articulated motion.

Semantic segmentation based on RGB and thermal images is an effective way to achieve an all-day understanding of road scenes. However, how to fuse the RGB and thermal information effectively remains an open problem. By studying fusion strategies at different stages, an illumination-aware feature fusion network has been developed for all-day semantic segmentation of urban road scenes in this paper, called IAFFNet. At the encoding stage, we introduce a bi-directional guided feature fusion module to effectively recalibrate and unify both RGB and thermal information. At the decoding stage, we have developed adaptive fusion modules to fuse low-level details and high-level semantic information. Finally, we have developed a decision-level illumination-aware strategy to achieve robust all-day segmentation. As far as we know, we are the first to incorporate illumination clues explicitly into RGB-T semantic segmentation. Extensive experimental evaluations demonstrate that the developed method can achieve remarkable performance on public datasets compared with state-of-the-art methods.

In this paper, our focus is to segment the foreground area for human detection. It is assumed that the foreground region has been detected. Accurate foreground contours are not required. The developed approach adopts a modified ISM (Implicit Shape Model) to collect some typical local patches of human being and their location information. Individuals are detected by grouping some local patches in the foreground area. The method can get good results in crowded scenes. Some examples based on CAVIAR dataset have been shown. A main contribution of the paper is that ISM model and joint occlusion analysis are combined for individual segmentation. There are mainly two advantages: First, with more sufficient information inside the foreground region, even the individuals inside a dense area can also be handled. Secondly, the method does not require an accurate foreground contour. A rough foreground area can be easily obtained in most situations.

The cross-view image translation task is aimed at generating scene images from arbitrary views. However, due to the great differences in the shapes and contents of the various views, the quality of the generated images is degraded. Small objects, such as vehicles’ shapes and details, are not clear, which causes them to be structurally inconsistent with the semantic map used to guide the generation process. To solve this problem, we propose a novel generative adversarial network based on a local and global information processing module (LAGGAN) to recover the image’s details and structures. The network will further combine the input viewpoint image and the target semantic segmentation map to guide the generation of the target image from another viewpoint. The proposed LAGGAN includes a two-stage generator and a parameter-sharing discriminator. LAGGAN uses a new local and global information processing module (LAG) to generate high-quality images from various views. Moreover, we integrate dilated convolutions into the discriminator to capture the global context, which can enhance the discriminative ability and further adjust the LAG module. Therefore, most semantic information can be preserved, and the details of the target viewpoint images can be translated more sharply. Quantitative and qualitative evaluation on both CVUSA and Dayton datasets attest to the fact that our method, LAGGAN, presents satisfactory perceptual results and is comparable to state-of-the-art methods on the cross-view image translation task.