Object detection in drone-captured images is a popular task recent years. As drones always navigate at different altitudes, the object scale varies considerably, which burdens optimization of models. Moreover, high-speed and low-altitude flight cause motion blur on densely packed objects, leads to great challenges. To solve two issues mentioned above, based YOLOv5, we add an additional prediction head detect tiny-scale objects replace CNN-based heads with transformer (TPH), constructing...

Semantic segmentation is a critical module in robotics related applications, especially autonomous driving. Most of the research on semantic focused improving accuracy with less attention paid to computationally efficient solutions. Majority algorithms have customized optimizations without scalability and there no systematic way compare them. In this paper, we present real-time benchmarking framework study various for We implemented generic meta-architecture via decoupled design where...

Inspired by the recent advances of image super-resolution using convolutional neural network (CNN), we propose a CNN-based block up-sampling scheme for intra frame coding. A can be down-sampled before being compressed normal coding, and then up-sampled to its original resolution. Different from previous studies on down/up-sampling-based methods in our have been designed training CNN instead hand-crafted. We explore new structure up-sampling, which features deconvolution feature maps,...

Aiming at simultaneous detection and segmentation (SD-S), we propose a proposal-free framework, which detect segment object instances via mid-level patches. We design unified trainable network on patches, is followed by fast effective patch aggregation algorithm to infer instances. Our method benefits from end-to-end training. Without proposal generation, computation time can also be reduced. In experiments, our yields results 62.1% 61.8% in terms of mAPr VOC2012 val SDS val, are...

Semantic segmentation for high-resolution remote sensing images is one of the most significant tasks in field applications. Remote contain substantial detailed information ground objects, such as shape, location, and texture. Therefore, these objects make exhibit large intraclass variance small interclass variance, which makes it very difficult to be recognized. In this study, an end-to-end attention-based semantic network (SSAtNet) proposed. A pyramid attention pooling module proposed...

According to Darwinian evolutionary theory, numerous species in the wild have developed remarkable adaptive mechanisms, involving pattern rearrangement and environmental assimilation, evade predators. These obfuscation strategies pose significant challenges for both individuals algorithms when performing Camouflage Object Detection (COD) task complex intricate scenarios. Inspired by human COD task, which involve assigning uncertainties entire input then focusing on highly uncertain areas...

Recently, by virtue of the high computational efficiency and accuracy, discriminative correlation filter (DCF)- based tracking methods have gained attraction in field unmanned aerial vehicle (UAV). However, conventional DCF-based merely rely on cyclic shift to produce training samples. As a result, trained these samples owns limited ability, ineffectively addressing various challenges stage. Here, promote filter's we develop feature block-aware (CF) method. Specifically, extracted is divided...

With the wide applications of vision based intelligent systems, image and video analysis technologies have attracted attention researchers in computer field. In analysis, human activity recognition is an important research direction. By interpreting understanding activity, we can recognize predict occurrence crimes help police or other agencies react immediately. past, a large number papers been published on sequences. this paper, provide comprehensive survey recent development techniques,...

Timely and accurately detecting personal protective equipment (PPE) usage among workers is essential for substation safety management. However, traditional algorithms encounter difficulties in substations due to issues such as varying target scales, intricate backgrounds, many model parameters. Therefore, this paper proposes MEAG-YOLO, an enhanced PPE detection built upon YOLOv8n. First, the incorporates Multi-Scale Channel Attention (MSCA) module improve feature extraction. Second, it newly...

As the Internet-of-Things (IoT) and edge computing have been major paradigms for distributed data collection, communication, processing, smart city applications in real world tend to adopt IoT broadly. Today, more machine learning algorithms would be deployed into front-end sensors, devices, centres rather than centralised cloud centres. However, sensors devices are usually not so capable as those units huge centres, this sake, practice, engineers choose compromise limited capacity of...

As a significant research direction in remote sensing fields, unmanned aerial vehicles (UAVs) tracking has achieved rapid development recent years. However, due to limited power and computation resources on platforms, the methods deployed UAVs usually require high computational efficiency performance. In addition, various challenges (i.e., similar object, background clutter occlusion) have inevitably occurred during UAV phase. Therefore, considering above issues comprehensively, this paper...

This paper proposes a novel cross-media retrieval approach. First, an isomorphic subspace is constructed based on canonical correlation analysis (CCA) to learn multi-modal correlations of media objects; second, polar coordinates are used judge the general distance objects with different modalities in subspace. Since integrity semantic not likely learned from limited training samples, users' relevance feedback accurately refine similarities. We also propose methods map new into subspace, and...

Abstract Objective Current approaches to food volume estimation require the person carry a fiducial marker (e.g. checkerboard card), be placed next before taking picture. This procedure is inconvenient and post-processing of picture time-consuming sometimes inaccurate. These problems keep people from using smartphone for self-administered dietary assessment. The current bioengineering study presents novel smartphone-based imaging approach table-side which overcomes limitations. Design We...

In recent years, the excellent image-based object detection algorithms are transferred to video directly. These frame-by-frame processing methods suboptimal owing degenerate appearance such as motion blur, defocus and rare poses. The existing works for mostly focus on feature aggregation at pixel level instance level, but blur impact in process has not been exploited well so far. this article, we propose an end-to-end blur-aid network (BFAN) detection. proposed BFAN focuses influenced by...

The rapid development of pay-as-you-go cloud services motivates the increasing number resource demands. However, volatile demands bring new challenges for current techniques to minimize cost capacity planning and VM provisioning while satisfying customer service vendors will incur enormous revenue loss within long-term inappropriate planning, especially when fluctuate abruptly frequently. In this paper, we cast as a classification problem propose an integrated framework, which effectively...

To address the issues of fuzzy scene details, reduced definition, and poor visibility in images captured under non-uniform lighting conditions, this paper presents an algorithm for effectively enhancing such images. Firstly, adaptive color balance method is employed to differences low-light images, ensuring a more uniform distribution yielding image with improved consistency. Subsequently, obtained transformed from RGB space HSV space, wherein multi-scale Gaussian function utilized...