A5101767499- Domain Adaptation and Few-Shot Learning
- Advanced Image and Video Retrieval Techniques
- Advanced Neural Network Applications
- Human Pose and Action Recognition
- Advanced Image Processing Techniques
- Multimodal Machine Learning Applications
- Plant Micronutrient Interactions and Effects
- Soil Carbon and Nitrogen Dynamics
- Radiomics and Machine Learning in Medical Imaging
- Agricultural Science and Fertilization
- Soil and Water Nutrient Dynamics
- Agriculture, Soil, Plant Science
- Crop Yield and Soil Fertility
- Anomaly Detection Techniques and Applications
- Infrared Target Detection Methodologies
- COVID-19 diagnosis using AI
- Insect Resistance and Genetics
- Protein Structure and Dynamics
- AI in cancer detection
- Esophageal Cancer Research and Treatment
- Gait Recognition and Analysis
- Advanced Image Fusion Techniques
- Generative Adversarial Networks and Image Synthesis
- Sparse and Compressive Sensing Techniques
- Dengue and Mosquito Control Research
Xinjiang Uygur Autonomous Region Disease Prevention and Control Center
2022-2024
Xinjiang Medical University
2022-2024
Fuyang Normal University
2015-2022
Guangzhou Sport University
2021
Jiangxi Academy of Agricultural Sciences
2010-2020
Institute of Automation
2019
Chinese Academy of Sciences
2013-2019
Hunan Academy of Agricultural Sciences
2018
Group Sense (China)
2016
University of Chinese Academy of Sciences
2014-2015
The central building block of convolutional neural networks (CNNs) is the convolution operator, which enables to construct informative features by fusing both spatial and channel-wise information within local receptive fields at each layer. A broad range prior research has investigated component this relationship, seeking strengthen representational power a CNN enhancing quality encodings throughout its feature hierarchy. In work, we focus instead on channel relationship propose novel...
We present a state-of-the-art image recognition system, Deep Image, developed using end-to-end deep learning. The key components are custom-built supercomputer dedicated to learning, highly optimized parallel algorithm new strategies for data partitioning and communication, larger neural network models, novel augmentation approaches, usage of multi-scale high-resolution images. Our method achieves excellent results on multiple challenging computer vision benchmarks.
While the use of bottom-up local operators in convolutional neural networks (CNNs) matches well some statistics natural images, it may also prevent such models from capturing contextual long-range feature interactions. In this work, we propose a simple, lightweight approach for better context exploitation CNNs. We do so by introducing pair operators: gather, which efficiently aggregates responses large spatial extent, and excite, redistributes pooled information to features. The are cheap,...
Recently, deep learning approaches have demonstrated remarkable progresses for action recognition in videos. Most existing frameworks equally treat every volume i.e. spatial-temporal video clip, and directly assign a label to all volumes sampled from it. However, within video, discriminative actions may occur sparsely few key volumes, most other are irrelevant the labeled category. Training with large proportion of will hurt performance. To address this issue, we propose mining framework...
The sparse coding technique has shown flexibility and capability in image representation analysis. It is a powerful tool many visual applications. Some recent work that incorporating the properties of task (such as discrimination for classification task) into dictionary learning effective improving accuracy. However, traditional supervised methods suffer from high computation complexity when dealing with large number categories, making them less satisfactory scale In this paper, we propose...
For the task of visual categorization, learning model is expected to be endowed with discriminative feature representation and flexibilities in processing many categories. Many existing approaches are designed based on a flat category structure, or rely set pre-computed features, hence may not appreciated for dealing large numbers In this paper, we propose novel dictionary method by taking advantage hierarchical correlation. each internode classification models learnt dictionaries different...
Ovarian cancer is a highly lethal malignancy in the field of oncology. Generally speaking, segmentation ovarian medical images necessary prerequisite for diagnosis and treatment planning. Therefore, accurately segmenting tumors utmost importance. In this work, we propose hybrid network called PMFFNet to improve accuracy tumors. The utilizes an encoder-decoder architecture. Specifically, encoder incorporates ViTAEv2 model extract inter-layer multi-scale features from feature pyramid. To...
Abstract The response of soil microbial communities to quality changes is a sensitive indicator ecosystem health. current work investigated under different fertilization treatments in 31-year experiment using the phospholipid fatty acid (PLFA) profile method. consisted five treatments: without fertilizer input (CK), chemical alone (MF), rice ( Oryza sativa L.) straw residue and (RF), low manure rate (LOM), high (HOM). Soil samples were collected from plough layer results indicated that...
Multi-task learning has been proposed to improve the generalization performance by multiple tasks jointly. One challenge for this paradigm is effectively seek shared information across tasks. In paper, we propose a novel multi-task method adaptively share information. Unlike many existing methods which impose strong assumptions on task related-ness, our captures relationships among and identifies disparities of each simultaneously, thus can flexibly exploit Moreover, apply it fine-grained...
Super-resolution reconstruction has been widely used in infrared images. A lot of effective super-resolution methods have presented recent years. In this paper, a fast and robust algorithm based on Maximum Posteriori (MAP) estimation is proposed to obtain high resolution image from set images, which are obtained by an uncooled detector. comparison analysis made the results method, with variance regularizations number low direct observation value Power Signal-to-Noise Ratio (PSNR). Simulation...
Weakly-supervised object detection (WSOD) has attracted lots of attention in recent years. However, there is still a big gap between WSOD and generic detection. The main barriers to the efficiency are ineffective data augmentations inaccurate bounding box predictions. Given only image-level annotations, it hard for effectively utilize variant accurately regress boxes. Although fully-supervised detector can be trained using annotations generated from weakly-supervised detector, performance...
Hierarchical classification models have been proposed to achieve high accuracy by transferring effective information across the categories. One important challenge for this paradigm is design what can be transferred In paper, we propose a novel method learn sharing model taking advantage of multi-level feature representations. Unlike many existing methods which based on identical space, detectors enable our capture rich visual in hierarchical category structure. Moreover, classifier...