A tomato automatic detection method based on an improved YOLOv8s model is proposed to address the low automation level in harvesting agriculture. The provides technical support for and classification of tomatoes agricultural production activities. has three key components. Firstly, depthwise separable convolution (DSConv) technique replaces ordinary convolution, which reduces computational complexity by generating a large number feature maps with small amount calculation. Secondly, dual-path...

Skeleton-based action recognition has attracted considerable attention since the skeleton data is more robust to dynamic circumstances and complicated backgrounds than other modalities. Recently, many researchers have used Graph Convolutional Network (GCN) model spatial-temporal features of sequences by an end-to-end optimization. However, conventional GCNs are feedforward networks for which it impossible shallower layers access semantic information in high-level layers. In this paper, we...

Convolutional Neural Networks have achieved excellent successes for object recognition in still images. However, the improvement of over traditional methods recognizing actions videos is not so significant, because raw usually much more redundant or irrelevant information than In this paper, we propose a Spatial-Temporal Attentive Network (STA-CNN) which selects discriminative temporal segments and focuses on informative spatial regions automatically. The STA-CNN model incorporates Temporal...

Recommending fashion outfits to users presents several challenges. First of all, an outfit consists multiple items, and each user emphasizes different parts when considering whether they like it or not. Secondly, a user's liking for considers not only the aesthetics item but also compatibility among them. Lastly, data is often sparse in terms relationship between outfits. Not mention, we can obtain what like, dislike.

The accurate localization of S1 and S2 is essential for heart sound segmentation classification. However, current direct algorithms have poor noise immunity low accuracy. Therefore, this paper proposes a new optimal algorithm based on K-means clustering Haar wavelet transform. includes three parts. Firstly, method uses the Viola integral Shannon’s energy-based to extract function envelope energy. Secondly, time–frequency domain features acquired are extracted from different dimensions peak...

Multi-task learns multiple tasks, while sharing knowledge and computation among them. However, it suffers from catastrophic forgetting of previous when learned incrementally without access to the old data. Most existing object detectors are domain-specific static, some but only within a single domain. Training an detector across various domains has rarely been explored. In this work, we propose three incremental learning scenarios categories for detection. To mitigate forgetting, attentive...

Convolutional Neural Network (CNN) and Recurrent (RNN) are two typical kinds of neural networks. While CNN models have achieved great success on image recognition due to their strong abilities in abstracting spatial information from multiple levels, RNN not significant progress video analyzing tasks (e.g. action recognition), although can inherently model temporal dependencies videos. In this work, we propose a Sequential Network, denoted as SCNN, extract effective spatial-temporal features...

Knowledge graphs (KGs) are of great importance to many artificial intelligence applications, but they usually suffer from the incomplete problem. graph embedding (KGE), which aims represent entities and relations in low-dimensional continuous vector spaces, has been proved be a promising approach for KG completion. Traditional KGE methods only concentrate on structured triples, while paying less attention type information entities. In fact, incorporating entity types into learning could...

Summary In flowering plants, male germline fate is determined after asymmetric division of the haploid microspore. Daughter cells have distinct fates: generative cell ( GC ) undergoes further mitosis to generate sperm SC s), and vegetative VC terminally differentiates. However, our understanding mechanisms underlying development remains limited. Histone variants modifications define chromatin states, contribute establishing maintaining identities by affecting gene expression. Here, we...

Fashion is more than Paris runways. about how people express their interests, identity, mood, and cultural influences. Given an inventory of candidate garments from different categories, to assemble them together would most improve fashionability? This question presents intriguing visual recommendation challenge automatically create capsule wardrobes. Capsule wardrobe generation a complex combinatorial problem that requires the understanding multiple items interact. The generative process...

Fashion-on-demand is becoming an important concept for fashion industries. Many attempts have been made to leverage machine learning methods generate designs tailored customers' tastes. However, how assemble items together (e.g., compatibility) crucial in designing high-quality outfits synthesis images. Here we propose a generation model, named OutfitGAN, which contains two core modules: Generative Adversarial Network and Compatibility Network. The generative module able new realistic high...

Skeleton-based action recognition has attracted considerable attention in computer vision since skeleton data is more robust to the dynamic circumstance and complicated background than other modalities. Recently, many researchers have used Graph Convolutional Network (GCN) model spatial-temporal features of sequences by an end-to-end optimization. However, conventional GCNs are feedforward networks which impossible for low-level layers access semantic information high-level layers. In this...

Existing knowledge distillation (KD) methods are mainly based on features, logic, or attention, where features and logic represent the results of reasoning at different stages a convolutional neural network, attention maps symbolize process. Because continuity two in time, transferring only one them to student network will lead unsatisfactory results. We study transfer between teacher-student degrees, revealing importance simultaneously related process providing new perspective for KD. On...

Compressed sensing (CS) MRI has shown great potential in enhancing time efficiency. Deep learning techniques, specifically generative adversarial networks (GANs), have emerged as potent tools for speedy CS-MRI reconstruction. Yet, the complexity of deep reconstruction models increases, this can lead to prolonged and challenges achieving convergence. In study, we present a novel GAN-based model that delivers superior performance without escalating. Our generator module, built on U-net...

Video anomaly detection is an essential and challenging task in the computer vision community, which aims to automatically detect localize abnormal events videos. In this paper, we propose attention augmented spatial-temporal normality learning framework explore unique spatial temporal patterns of normal events. Specifically, first slice videos into local cubes along dimensions facilitate independent prototypical training phase, use parallel deep convolutional neural networks learn features...