<title>Abstract</title> Face detection and landmark localization are usually the fundamental important steps of facial analysis. The accuracy degradation any task has impacts on robustness downstream In most analysis systems, face detection, as two independent tasks, predicted sequentially with single-task detectors respectively, which suffer from higher complexity poorer consistency. contrast, multi-task joint more efficient have shared features. However, when tasks in one model, it is...

To solve the complex background problems (e.g. Noise interference, object overlap, and different illumination) that affect classification performance on plant leaf images, this paper proposes an instance segmentation method for images based IFPN SNMS CFFI-Mask R-CNN (ISC-MRCNN) ACPSOSVM-Dual Channels Convolutional Neural Network (APS-DCCNN). obtain foreground of lateral connection structure feature map pyramid in ISC-MRCNN fuses maps depths, so network learns more detailed features. Then,...

Salicylic acid (SA) is a natural inducer of disease resistance in fruit, but its application the food industry limited due to low water solubility. Here, SA was encapsulated β-cyclodextrin (β-CD) via host–guest inclusion complexation method, and efficacy microcapsules (SAM) against blue mold caused by Penicillium expansum postharvest apple fruit elucidated. It observed that SAM most effective inhibiting mycelial growth P. vitro. also superior for control under vivo conditions. Enzyme...

As the only forests situated at transition between land and sea, mangrove are one of first ecosystems vulnerable to rising sea levels. When level rises, plants exposed increased salinity, as well tidal flooding. The responses changing levels depend on synergistic effects flooding salinity plants, especially seedlings. focus this paper is assess ability different tide position Aegiceras corniculatum (A. corniculatum) Bruguiera sexangula (B. sexangula) seedlings withstand seawater investigate...

Machine learning (ML) requires the collection of large amounts data to train robust predictive models. Federated (FL) allows sensitive be kept on client side a shared model, but models still pose privacy concerns. Secure aggregation is an important algorithm for securing federation by being able compute sum client-side model updates without revealing information about individual updates. In this study, we investigate results secure protocols in recent years and review research field...

High-resolution radar imaging will give us detailed information of target, which becomes basic function systems. Improvement image resolution the existing system is also important. Based on deep learning, a new method for super-resolution through-the-radar proposed. A network, called cascade U-net (CU-net), proposed in this paper. The results simulation and real data experiments demonstrate effectiveness our methods.

In this paper, a method of polarimetric SAR image classification based on decomposition and AdaBoost algorithm is proposed. The proposed improves accuracy speed. algorithm, as robust learner with high accuracy, can fully utilize the features to achieve classification. simulated tests, observed produce improved speed, compared H /α̅ algorithm.

Abstract A distinctive and intricate color palette is often used for interior decorations in Xizang temples. Traditional methods of detection can be challenging, as they are prone to being limited by similar background colors, making it difficult identify all the essential elements. In this case, a momentum-based pseudo-labeling strategy proposed address issue. This utilizes momentum expansion categorize pseudo-labels generated first-stage pre-trained model into active areas, facilitating...

Abstract Revealing the impact of forest succession processes on changes in plant diversity is crucial for understanding mechanisms that maintain across various stages. While previous research has predominantly focused influence environmental factors or management strategies within rubber plantation understories, there a scarcity studies examining effects diversity. This study focuses understory herbaceous layer Yinggeling area, located National Park Hainan Tropical Rainforest. It employs...

As a 3-D point cloud has the ability to present contour of an object clearly, it provides more spatial information for person identification (PI) task. Aiming at improvements on quality and distribution features, innovative treatment method novel network structure are investigated in this article. First, spatiotemporal feature is enhanced by implementing dual-stage density-based clustering applications with noise (DST-DBSCAN) method, which can filter most invalid points decrease sparsity...

This work aims to leverage pre-trained foundation models, such as contrastive language-image pre-training (CLIP) and segment anything model (SAM), address weakly supervised semantic segmentation (WSSS) using image-level labels. To this end, we propose a coarse-to-fine framework based on CLIP SAM for generating high-quality seeds. Specifically, construct an image classification task seed task, which are jointly performed by with frozen weights two sets of learnable task-specific prompts. A...

Current intelligent diagnosis systems struggle to continually learn diagnose more and diseases due catastrophic forgetting of old knowledge when learning new knowledge. Although storing small data for subsequent continual can effectively help alleviate the issue, heavy imbalance between classes to-be-learned in classifier training often causes biased prediction towards just learned by updated classifier. In this study, an outlier detection technique is novelly applied train additional expert...

Decision-making for multi-snip collision avoidance usually concerns people's experience and judgment, so it is difficult to make a pure quantificational model. An adaptive genetic algorithm model proposed in this paper. It introduces kind of encoding method combining binary numbers decimal numbers. In the solutions space consists own ship's course speed. For improving performance speed guaranteeing precision, integer part represented with 9 numbers; 1 number. The 7 Finally, genome string,...

Although decision tree learning has achieved great success in building classifier, most existing methods don't pay attention to unequal weights between different instances from training and testing data sets. However, many real world sets are imbalanced nature. In this paper, we introduce a new improved based weights, which considers instances, address the class problems. The proposed algorithm is simple more effective implementation than previous trees. Also, will be compared with C4.5 (a...

Light, soil temperature, and moisture likely change when a gap or clearing is formed in closed-canopy forests, triggering seed germination the bank. However, which factors induce such remains elusive. In this study, we explored mechanism of gaps on bank without felling trees. Power sources were supplied forest, three types lamps used to simulate photo-thermal, light, temperature environments (i.e., photo-thermal separation field), respectively. The experiment was carried out tropical rain...

Analyze the indications of empirical prescription and its relations in TCM, provide a clear guide for application. [Methods]: First, we collect cases that professor Zhi-Min Yang uses to remedy insomnia investigate patient's demographic feature, historical four diagnostic techniques TCM Pittsburgh Sleep Quality Index(PSQI). Second, observe whole treatment, define standard curative effect by PSQI reduced rate between before-treatment post-treatment 1 month. Third, divide into invalid group...