Camouflaged object detection aims to detect objects closely blended into the background. Inspired by visual mechanism of human being, we define detection process as an end-to-end task consisted of locating and refining. To this end, we propose Contour Supervision and Initial Locating Guidance Network (CSIGNet) to effectively segment camouflaged objects from background. Specifically, our method fully explores the contribution of semantic contour to the binary segmentation task. In addition, attention mechanism is used for our final prediction. Experiments show that it has achieved excellent results on public datasets and the model can accurately segment camouflage objects. Codes will be made available at https://github.com/RecKono/CSIGNet.

Fully convolutional neural networks have achieved great success in salient object detection, in which the effective use of multi-layer features plays a critical role. Based on this advantage, many saliency detectors have emerged in recent years, and most of them designed a series of network structures to integrate the multi-level features generated by the backbone network. However, information in different layer play different roles in saliency object detection, how to integrate them effectively is still a great challenge. In this article, a selective feature fusion network which consists of a selective feature fusion module (SFM) and an attention-guide hierarchical feature emphasis module (AEM) is proposed. Most of the previous works mainly integrate multi-level feature by addition and concatenation, as a difference, SFM adaptively selects the important information from the input features in the fusion, which effectively avoids introducing too much redundant information. Besides, AEM combines spatial attention and channel attention to enhance features simply and effectively by hierarchical iteration, and further improve the accuracy of salient object detection. Experiments on five datasets show that the proposed selective feature fusion method achieve satisfactory results when comparing to other state-of-the-art salient object detection approaches.

Fully convolutional neural networks-based salient object detection has recently achieved great success with its performance benefits from the effective use of multi-layer features. Based on this, most of the existing saliency detectors designed complex network structures to fuse the multi-level features generated by the backbone network. However, the variable scale and complex shape of the target are always a great challenge for saliency detection tasks. In this paper, the authors propose a Rich-scale Feature Fusion Network (RFFNet) for salient object detection. The authors design a rich-scale feature interactive fusion module to obtain more efficient features from the multi-scale features. Moreover, the global feature enhance module is used to extract features with better characterization for the final saliency prediction. Extensive experiments performed on five benchmark datasets demonstrate that the proposed method can achieve satisfactory results on different evaluation metrics compared to other state-of-the-art salient object detection approaches.

Cross-modal information fusion plays a vital role in the RGB-T salient object detection. Due to RGB and thermal images come from different domains, the modality difference will lead to the unsatisfactory effect of simple feature fusion. How to explore and integrate useful information is the key to the RGB-T saliency detection methods. In this paper, we introduce an Enhancement and Fusion Network. In detail, we propose a Self-modality Feature Enhancement Module that effectively integrate the feature representation of a single modality through global context information. And we propose a Cross-modality Feature Dynamic Fusion Module to realize the effective fusion of cross-modal features in the way of dynamic weighting. Experiments on public datasets show that the proposed method achieves satisfactory results compared with other state-of-the-art salient object detection approaches.

High quality instance segmentation has shown emerging significance in computer vision, especially in complex camouflaged situations. For this reason, we propose a single-stage segmentation model named Contour-assisted Long-range Perceptual Network (CLPNet). By following SOLOv2 model, based on deformable transformer, we propose Aggregate and Optimize multi-layer Transformer for generating refining features. Secondly, through iterative optimization, the full use of long-range context dependencies makes the internal of the instance give strong response, and contour of object tightly surrounds the instance. Experiments on camouflaged object detection dateset show that our method reaches 41.7% AP. Compared with the previous instance segmentation method, it is obviously more effective in small object detection.

目前视频目标分割算法多是基于匹配和传播策略分割目标,常常以掩模或者光流的方式利用前一帧的信息,探索了新的帧间特征传播方式,利用短时匹配模块提取前一帧信息并传播给当前帧,提出一种面向视频序列数据的目标分割模型.通过长时匹配模块和短时匹配模块分别与第一帧和前一帧做相关操作进行像素级匹配,得到的全局相似性图和局部相似性图,以及前一帧的掩模和当前帧的特征图,经过两个优化网络后通过分割网络得到分割结果.在视频目标分割公开数据集上的实验表明,所提出方法在单目标和多目标上分别取得了86.5％和77.4％的区域相似度和轮廓精度均值,每秒可计算21帧.提出的短时匹配模块比仅使用掩模更有利于提取前一帧的信息,通过长时匹配模块和短时匹配模块的结合,不使用在线微调即可实现高效的视频目标分割,适合应用于移动机器人视觉感知.

Semi-supervised video object segmentation (semi-VOS) predicts pixel-accurate masks of the target objects in all frames according to the ground truth mask provided in the first frame. A critical challenge to this task is how to model the dependency between the query frame and other frames. Most methods neglect or do not make full use of the inherent relevance. In this paper, we propose a novel network called CO-Attention Matching Network (COMatchNet) for semi-VOS. The COMatchNet mainly consists of a co-attention module and a matching module. The co-attention module extracts frame correlation among the query frame and the previous frame and the first frame. The matching module calculates pixel-level matching scores and finds the most similar regions to preceding frames in the query frame. The COMatchNet integrates two level information and generates fine-grained segmentation masks. We conduct extensive experiments on three popular video object segmentation benchmarks, i.e. DAVIS 2016; DAVIS 2017; YouTube-VOS. Our COMatchNet achieves competitive performance (J&F) of 86.8%, 75.9%, and 81.4% on the above benchmarks, respectively.

Saliency object detection has become an active topic in both computer vision and multimedia fields. Though much remarkable progress has been achieved, it is still challenging to predicted saliency maps correctly in complex scenes due to low contrast or huge similarity between salient object and background region. Previous methods only use pixels from foreground objects. As a difference, background pixels are used for salience detection in this research. Specifically, this paper introduce a foreground-background features separation module with the attention mechanism to extract this pair of complementary features. And a spatial feature refinement module was designed to learn the spatial refinement feature. Useing the spatial feature of the low-layer to enhance the foreground and background features, and combine them to generate a final saliency map. Extensive experiments performed on five benchmark datasets demonstrate that the proposed method can achieve satisfactory results on different evaluation metrics compared to other state-of-the-art salient object detection approaches.

Although the video surveillance system plays an important role in intelligent transportation, the limited camera views make it difficult to observe many traffic events. In this paper, we collect and combine the traffic flow variables from the multi-source sensors, and propose a PITED method based on Random Forest (RF) and Permutation importance (PI) for traffic event detection. This model selects the suitable traffic flow variables by means of permutation arrangement of importance, and establishes the whole process of acquisition, preprocessing, quantization, modeling and evaluation. Moreover, the real traffic data are collected and tested in this paper for evaluating the experiment performance, including the miss/false rate of traffic event, and average detection time. The experimental results show that the detection rate is more than 85% and the false alarm rate is less than 3%. It means the model is effective and efficient in the practical application regardless of both workdays and holidays.

Fully convolutional neural network-based salient object detection has recently achieved great success with its performance benefits from the effective use of multi-layer features. Based on this, most of the existing saliency detectors design complex network structures to fuse the multi-level features of the backbone feature network. However, information in different layer play different roles in saliency object detection, how to integrate them is still an open problem. In this paper, a cascaded bottom-up feature aggregation module is designed to retain and strengthen more spatial details in the low-level features, and embed attention mechanism in the process of feature aggregation to filter more effective features. Extensive experiments show that the proposed networks can consistently improve saliency detection performance. The experimental results on five public datasets prove that this network is competitive in saliency detection.