Recently, tracking-by-detection has become a popular paradigm in Multiple-object tracking (MOT) for its concise pipeline. Many current works first associate the detections to form track proposals and then score proposalns by manual functions select best. However, long-term information is lost this way due detection failure or heavy occlusion. In paper, Extendable Multiple Nodes Tracking framework (EMNT) introduced model association. Instead of detections, EMNT creates four basic types nodes...

Light field (LF) semantic segmentation is a newly arisen technology and widely used in many smart city applications such as remote sensing, virtual reality 3D photogrammetry. Compared with RGB images, LF images contain multi-layer contextual information rich geometric of real-world scenes, which are challenging to be fully exploited because the complex highly inter-twined structure LF. In this paper, Contextual Feature (LFCF) Geometric (LFGF) proposed respectively for occluded area...

The rapid increase in the volume of video data generated from edges Industrial Internet Things, opens up new possibilities for enhancing application service. Multicamera multiobject tracking (MCMT) has always been a fundamental task surveillance or traffic control. However, traditional MCMT methods are limited by communication bottleneck and computation resources centralized curator, suffer security privacy issues. In this article, we first design multicamera multihypothesis (MC-MHT)...

Recently, most multiple object tracking (MOT) algorithms adopt the idea of tracking-by-detection. Relevant research shows that performance detector obviously affects tracker, while improvement is gradually slowing down in recent years. Therefore, trackers using tracklet (short trajectory) are proposed to generate more complete trajectories. Although there various generation algorithms, fragmentation problem still often occurs crowded scenes. In this paper, we introduce an iterative...

In recent years, the demand for intelligent devices related to Internet of Things (IoT) is rapidly increasing. field computer vision, many algorithms have been preinstalled in IoT achieve higher efficiency, such as face recognition, area detection, target tracking, etc. Tracking an important but complex task that needs high efficiency solutions real applications. There a common assumption detection can only represent one pedestrian describe nonoverlapping physical space. fact, pixels image...

Multiobject tracking is a basic task in video analysis. Due to the strict requirements on efficiency and resource consumption, most of applications edge devices are online or near-online methods. Besides motion modeling, appearance information also widely used for tracking. However, influence occlusion usually ignored. In this article, spatial-temporal co-occurrence constraints (STCCs) features introduced resist occlusions by exploring rich spatial temporal tracklets. addition, novel...

Recent progress in multi-object tracking (MOT) has shown great significance of a robust scoring mechanism for potential tracks. However, the lack available data MOT makes it difficult to learn general mechanism. Multiple cues including appearance, motion and etc., are limitedly utilized current manual functions. In this paper, we propose Nodes Tracking (MNT) framework that adapts most trackers. Based on framework, Recurrent Unit (RTU) is designed score tracks through long-term information....

For an intelligent transportation system, multiple object tracking (MOT) is more challenging from the traditional static surveillance camera to mobile devices of Internet Things (IoT). To cope with this problem, previous works always rely on additional information multivision, various sensors, or precalibration. Only based a monocular camera, we propose hybrid motion model improve accuracy in devices. First, evaluates hypotheses by measuring optical flow similarity and transition smoothness...

Stereo vision is widely studied for depth information extraction. However, occlusion and noise pose significant challenges to traditional methods due failure in photo consistency. In this paper, an noise-aware stereo framework named ONAF proposed get a robust estimation by integrating the advantages of correspondence cues refocusing from light field(LF). consists two special cue extractors: extractor (CCE) (RCE). CCE extracts accurate areas based on multi-direction Ray-Epipolar Plane...

With the development of smart cities, video surveillance has become more prevalent in urban areas. The rapid growth data brings challenges to processing and analysis. Multi-object tracking (MOT), one most fundamental tasks computer vision, a wide range applications prospects. MOT aims locate multiple objects maintain their unique identities by analyzing frame frame. Most existing frameworks are deployed centralized systems, which convenient for management but have problems such as weak...

Breast cancer is the most commonly occurring worldwide. The ultrasound reflectivity imaging technique can be used to obtain breast (BUS) images, which classify benign and malignant tumors. However, classification subjective dependent on experience skill of operators doctors. automatic method assist doctors improve objectivity, but current convolution neural network (CNN) not good at learning global features vision transformer (ViT) extraction local features. In this study, we proposed a...

Vehicle Re-Identification (ReID) aims to find images of the same vehicle from different videos. It remains a challenging task in video analysis field due huge appearance discrepancy cross-view matching and subtle difference similar vehicles same-view matching. In this paper, we propose Co-occurrence Attention Net (CAN) deal with these two challenges. Specifically, CAN consists branches, main branch an aware branch. The is charge extracting global features that are consistent most views. This...

Compared to 2D imaging data, the 4D light field (LF) data retains richer scene's structure information, which can significantly improve computer's perception capability, including depth estimation, semantic segmentation, and LF rendering. However, there is a contradiction between spatial angular resolution during image acquisition period. To overcome above problem, researchers have gradually focused on super-resolution (LFSR). In traditional solutions, achieved LFSR based various...

Secure multi-party computation (MPC) has recently become prominent as a concept to enable multiple parties perform privacy-preserving machine learning without leaking sensitive data or details of pre-trained models the other parties.Industry and community have been actively developing promoting high-quality MPC frameworks (e.g., based on TensorFlow PyTorch) usage MPC-hardened models, greatly easing development cycle integrating deep with primitives.Despite prosperous adoption frameworks,...

To improve the generalization of autonomous driving (AD) perception model, vehicles need to update model over time based on continuously collected data. As progresses, amount data fitted by AD expands, which helps substantially. However, such ever-expanding is a double-edged sword for model. Specifically, as volume grows exceed model's fitting capacities, prone under-fitting. address this issue, we propose use pretrained Large Vision Models (LVMs) backbone coupled with downstream head...

Data association is one of the key research in tracking-by-detection framework. Due to frequent interactions among targets, there are various relationships trajectories crowded scenes which leads problems data association, such as ambiguity, omission, etc. To handle these problems, we propose hypothesis-testing based tracking (HTBT) framework build potential associations between target by constructing and testing hypotheses. In addition, a spatio-temporal interaction graph (STIG) model...

As online trade and interactions on the internet are rise, a key issue is how to use simple effective evaluation methods accomplish trust decision-making for customers. It well known that subjective holds uncertainty like randomness fuzziness. However, existing approaches which commonly based probability or fuzzy set theory can not attach enough importance uncertainty. To remedy this problem, new quantifiable approach proposed cloud model. Subjective modeled with model in approach, expected...

Multi-object tracking (MOT) has become a hot task in multi-media analysis. It not only locates the objects but also maintains their unique identities. However, previous methods encounter failures complex scenes, since they lose most of attributes each target. In this paper, we formulate MOT problem as Tracking Game and propose Self-adaptative Agent Tracker (SAT) framework to solve problem. The roles are divided into two classes including agent player game organizer. organizer controls...