Modern CNN-based object detectors assign anchors for ground-truth objects under the restriction of object-anchor Intersection-over-Unit (IoU). In this study, we propose a learning-to-match approach to break IoU restriction, allowing match in flexible manner. Our approach, referred as FreeAnchor, updates hand-crafted anchor assignment "free" matching by formulating detector training maximum likelihood estimation (MLE) procedure. FreeAnchor targets at learning features which best explain class...

Detecting oriented and densely packed objects remains challenging for spatial feature aliasing caused by the intersection of reception fields between objects. In this paper, we propose a convex-hull adaptation (CFA) approach configuring convolutional features in accordance with object layouts. CFA is rooted representation, which defines set dynamically predicted points guided convex over union (CIoU) to bound extent pursues optimal assignment constructing sets splitting positive or negative...

Modern CNN-based object detectors assign anchors for ground-truth objects under the restriction of object-anchor Intersection-over-Union (IoU). In this study, we propose a learning-to-match (LTM) method to break IoU restriction, allowing match in flexible manner. LTM updates hand-crafted anchor assignment "free" matching by formulating detector training Maximum Likelihood Estimation (MLE) framework. During phase, is implemented converting detection likelihood loss functions which are...

We present SegGPT, a generalist model for segmenting everything in context. unify various segmentation tasks into in-context learning framework that accommodates different kinds of data by transforming them the same format images. The training SegGPT is formulated as an coloring problem with random color mapping each sample. objective to accomplish diverse according context, rather than relying on specific colors. After training, can perform arbitrary images or videos via inference, such...

We present SegGPT, a generalist model for segmenting everything in context. unify various segmentation tasks into in-context learning framework that accommodates different kinds of data by transforming them the same format images. The training SegGPT is formulated as an coloring problem with random color mapping each sample. objective to accomplish diverse according context, rather than relying on specific colors. After training, can perform arbitrary images or videos via inference, such...

Abstract The advantages of the triboelectric nanogenerator (TENG) in environmental energy harvesting and utilization determine that they have great development potential digital application power grids. This work focuses on vibration from transmission lines, specifically context breeze vibrations sub‐span oscillations. To this end, proposes a novel multi‐mode vibrational (MV‐TENG) three smart line strategies based MV‐TENG. MV‐TENG design consists an S‐beam to broaden frequency response range...

This paper is aimed at solving the problem of poor room temperature plasticity 7B04 aluminum alloy T74 state. In this study, a method rapid heat treatment by laser scanning was adopted. Mechanical properties coupled with related microstructure evolution after were analyzed simultaneously. The results show that large impurity phase where microcracks usually generate in material state can be eliminated under appropriate parameters treatment. At same time, number fine strengthening phases...

Despite of the substantial progress visual object detection, models trained in one video domain often fail to generalize well others due change camera configurations, lighting conditions, and person views. In this paper, we present Domain Contrast (DC), a simple yet effective approach inspired by contrastive learning for training adaptive detectors. DC is deduced from error bound minimization perspective transferred model, implemented with cross-domain contrast loss which plug-and-play. By...

Detecting oriented and densely packed objects is a challenging problem considering that the receptive field intersection between causes spatial feature aliasing. In this paper, we propose convex-hull adaptation (CFA) approach, with aim to configure convolutional features in accordance irregular object layouts. CFA roots representation, which defines set of dynamically sampled points guided by convex over union (CIoU) bound extent. pursues optimal assignment constructing sets iteratively...

Modern object detectors have taken the advantages of backbone networks pre-trained on large scale datasets. Except for networks, however, other components such as detector head and feature pyramid network (FPN) remain trained from scratch, which hinders generalization capacity detectors. In this study, we propose to integrally migrate transformer encoder-decoders (imTED) a detector, constructing extraction path is "fully pre-trained" so that detectors' maximized. The essential differences...

Abstract Monitoring and controlling the vibration of bridge cables can maintain health structure prolong service life. This work proposes a novel self‐powered intelligent damper (SID) integrated triboelectric‐electromagnetic hybrid unit for in situ monitoring stay cables. Among them, sensing function SID is realized by an arched membrane triboelectric nanogenerator (AM‐TENG), energy harvesting three‐phase electromagnetic generator (TP‐EMG). The working principle performance are investigated....

Abstract Driver's driving behavior and style have a crucial impact on traffic safety, capacity, efficiency, so it is of great significance to monitor the driver's recognize their style. In this work, an intelligent monitoring system based triboelectric nanogenerator electromagnetic generator designed. The consists self‐powered pedal motion sensor (SPMS) data processing unit (IDPU), which can SPMS used for monitoring, mainly six‐phase (S‐TENG) free‐rotating disk (FD‐EMG). S‐TENG information...

Abstract Dynamic loads are inevitably generated during the operation of rotating machinery. In situ monitoring dynamic is great significance for assessing shafting health. Triboelectric nanogenerator (TENG) sensitive to motion but not suitable load monitoring, while piezoelectric (PENG) just opposite. Therefore, combination two can complement each other. From this, a smart bearing (DLMSB) integrated with TENG and PENG proposed. The specially designed produce an “M‐waveform”, which be used...

Abstract Triboelectric nanogenerator (TENG), as a new technique for energy capture, has unique advantages in harvesting high‐entropy energy. In view of the wind‐induced vibration characteristics power transmission lines, symmetric elastic bistable triboelectric (EB‐TENG) is introduced this work. The core component EB‐TENG cantilever beam structure, which designed by introducing an perturbation structure to conventional beam. This results several sub‐resonance frequencies addition natural...

Nickel-rich cathode materials have attracted widespread interest due to their high capacity; however, the structure is prone degradation and collapse during cycling, resulting in poor stability performance safety, hindering development of high-nickel materials. Here, we propose a straightforward method consume oxygen on surface primary particles high-temperature calcination process precursors, inducing coupled rearrangement cations, situ generation nano-sized mixed spinel/rock salt defect...

Recently, masked image modeling (MIM) has offered a new methodology of self-supervised pre-training vision transformers. A key idea efficient implementation is to discard the patches (or tokens) throughout target network (encoder), which requires encoder be plain transformer (e.g., ViT), albeit hierarchical transformers Swin Transformer) have potentially better properties in formulating inputs. In this paper, we offer design named HiViT (short for Hierarchical ViT) that enjoys both high...

The stay cable is susceptible to external loads resulting in vibration. Aiming at the large variation of vibration amplitude cable, a damping structured triboelectric nanogenerator (DS‐TENG) for cables energy graded harvesting proposed. DS‐TENG can not only achieve energy, but also provide adjustable additional by utilizing friction force generated wavy center pole and moving block. With magnitude 24 mm as limit, divided into two order states. experimental results suggest that output 250 V...

Triboelectric Nanogenerators In article number 2302353, Hengyu Li, Zhong Lin Wang, Tinghai Cheng and co-workers propose a multi-mode vibrational triboelectric nanogenerator (TENG) three smart grid application strategies, which achieve broadband vibration energy harvesting of breeze sub-span oscillation transmission lines. The strategies provide more complete implementation scheme for the digital TENGs in

The weight sharing scheme and spatial pooling operations in Convolutional Neural Networks (CNNs) introduce semantic correlation to neighboring pixels on feature maps therefore deteriorate their pixel-wise classification performance. In this paper, we implement an Orthogonal Decomposition Unit (ODU) that transforms a convolutional map into orthogonal bases targeting at de-correlating features. theory, complete decomposition produces which can perfectly reconstruct any binary mask...