A5100771541- Generative Adversarial Networks and Image Synthesis
- Advanced Image and Video Retrieval Techniques
- Advanced Neural Network Applications
- Multimodal Machine Learning Applications
- Domain Adaptation and Few-Shot Learning
- Advanced Image Processing Techniques
- Topic Modeling
- Computer Graphics and Visualization Techniques
- Natural Language Processing Techniques
- 3D Shape Modeling and Analysis
- Video Analysis and Summarization
- Robotic Path Planning Algorithms
- Image Processing Techniques and Applications
- Plant Stress Responses and Tolerance
- Software Engineering Research
- Image and Signal Denoising Methods
- Rice Cultivation and Yield Improvement
- Advanced Vision and Imaging
- Video Surveillance and Tracking Methods
- Plant Micronutrient Interactions and Effects
- Modular Robots and Swarm Intelligence
- Plant responses to water stress
- Face recognition and analysis
- Heavy metals in environment
- Digital Media Forensic Detection
Shihezi University
2022-2024
Ministry of Agriculture and Rural Affairs
2024
Jilin Agricultural University
2023-2024
China University of Geosciences
2024
Jinan University
2024
Northeast Agricultural University
2024
Hubei University Of Economics
2024
Zhejiang University
2024
Sichuan Agricultural University
2024
Ministry of Ecology and Environment
2024
We present a novel bird's-eye-view (BEV) detector with perspective supervision, which converges faster and bet-suits modern image backbones. Existing state-of-the-art BEV detectors are often tied to certain depth pretrained backbones like Vo Vn et, hindering the synergy between booming detectors. To address this limitation, we prioritize easing optimization of by introducing view supervision. end, propose two-stage detector; where proposals from head fed into bird' s-eye-view for final...
Recent progress in diffusion models has revolutionized the popular technology of text-to-image generation. While existing approaches could produce photorealistic high-resolution images with text conditions, there are still several open problems to be solved, which limits further improvement image fidelity and relevancy. In this paper, we propose ERNIE-ViLG 2.0, a large-scale Chinese model, progressively upgrade quality generated by: (1) incorporating fine-grained textual visual knowledge key...
Abstract Saline-sodic stress can limit the absorption of available zinc in rice, subsequently impacting normal photosynthesis and carbohydrate metabolism rice plants. To investigate impact exogenous application on grown saline-sodic soil, this study simulated conditions using two varieties, 'Changbai 9' 'Tonghe 899', as experimental materials. Rice seedlings at 4 weeks age underwent various treatments including control (CT), 2 μmol·L −1 treatment alone (Z), 50 mmol·L (S), with (Z + S). We...
Addressing the emotional needs of elderly in urban space design has increasingly become a vital concern. This study innovatively integrates theories with community outdoor spaces, thereby expanding research on categorization spaces. At 8 sites Yi Jie Qu, China, 330 residents were randomly recruited to assess their color responses (CER) landscapes these Based Affective Circumplex Model and Japanese Color Image Theory, Emotion was constructed visually represent overall tendencies significant...
Accumulation of geogenic phosphorus (P) in groundwater is an emerging environmental concern, which closely linked to coupled processes involving FeOOH and organic matter under methanogenic conditions. However, it remains unclear how P enrichment associated with methane cycling, particularly the anaerobic oxidation (AMO). This study conducted a comprehensive investigation carbon isotopes dissolved inorganic (DIC), CO2, CH4, alongside Fe isotopes, microbial communities, functions quaternary...
Saline-sodic stress restricts the absorption of zinc by rice, consequently impacting photosynthesis process rice plants. In this experiment, Landrace 9 was selected as test material and potting method employed to investigate influence ZnO nanoparticles (ZnO NPs) on chlorophyll fluorescence in grown saline-sodic land. The research findings demonstrate that application NPs proves be more advantageous for growth soil. Notably, significantly decreases levels Na
Conventional methods for the image-text generation tasks mainly tackle naturally bidirectional separately, focusing on designing task-specific frameworks to improve quality and fidelity of generated samples. Recently, Vision-Language Pre-training models have greatly improved performance image-to-text tasks, but large-scale pre-training text-to-image synthesis task are still under-developed. In this paper, we propose ERNIE-ViLG, a unified generative framework with transformer model. Based...
Diffusion-based text-to-image generation models like GLIDE and DALLE-2 have gained wide success recently for their superior performance in turning complex text inputs into images of high quality diversity. In particular, they are proven to be very powerful creating graphic arts various formats styles. Although current supported specifying style oil painting or pencil drawing, fine-grained features color distributions brush strokes hard specify as randomly picked from a conditional...
Automatic colorization of anime line drawing has attracted much attention in recent years since it can substantially benefit the animation industry. User-hint based methods are mainstream approach for colorization, while reference-based offer a more intuitive approach. Nevertheless, although improve feature aggregation reference image and drawing, results not compelling terms color consistency or semantic correspondence. In this paper, we introduce an attention-based model which channel-wise...
Soil salinization is a prevalent global environmental issue that significantly hampers crop growth and yield. However, there has been limited research on the impact of nitrogen fertilization various management practices in alleviating saline-sodic stress crops. In order to examine combined straw fertilizer application physiological photosynthetic characteristics rice paddy fields, three-year field experiment was conducted Jilin Province, China. The as split-zone trial, where main zone...
Software engineers working with the same programming language (PL) may speak different natural languages (NLs) and vice versa, erecting huge barriers to communication efficiency. Recent studies have demonstrated effectiveness of generative pre-training in computer programs, yet they are always English-centric. In this work, we step towards bridging gap between multilingual NLs PLs for large models (LLMs). We release ERNIE-Code, a unified pre-trained model 116 6 PLs. employ two methods...
Clothing segmentation and fine-grained attribute recognition are challenging tasks at the crossing of computer vision fashion, which segment entire ensemble clothing instances as well recognize detailed attributes products from any input human images. Many new models have been developed for in recent years, nevertheless accuracy is less than satisfactory case layered or fashion different scales. In this paper, a DEtection TRansformer (DETR) based method proposed to with high accuracy. model,...
In recent years, the burgeoning interest in diffusion models has led to significant advances image and speech generation. Nevertheless, direct synthesis of music waveforms from unrestricted textual prompts remains a relatively underexplored domain. response this lacuna, paper introduces pioneering contribution form text-to-waveform generation model, underpinned by utilization models. Our methodology hinges on innovative incorporation free-form as conditional factors guide waveform process...
Medical Image Super-Resolution plays a pivotal role in enhancing diagnostic accuracy. Transformer-based methods, such as Restoration Using Swin Transformer (SwinIR) and transformer for fast Magnetic Resonance Imaging (SwinMR), have shown prowess this area but also exhibit limitations. Specifically, LayerNorm channel normalization diminishes high-frequency detail, while the Multilayer Perceptron prioritizes global information over local information. Moreover, low-resolution inputs contain...
Recent Vision-Language Pre-trained (VLP) models based on dual encoder have attracted extensive attention from academia and industry due to their superior performance various cross-modal tasks high computational efficiency. They attempt learn representation using contrastive learning image-text pairs, however, the built inter-modal correlations only rely a single view for each modality. Actually, an image or text contains potential views, just as humans could capture real-world scene via...
In this study, we collected soybean inter-root soil (clay soil) from the cold region of Heilongjiang Province, China, screening for cold-tolerant phosphorus- and potassium-solubilizing bacteria by gradient-cooling-directed design mixed bacterial agents. This study screened phosphorus-solubilizing constructed We analyzed strain’s phosphorus/potassium solubilizing capacity, as well its organic acid secretion ability, to reveal mechanism detoxification phosphorus potassium. Clay Heilongjiang,...
Grass damage in the seedling corn field has always been an important factor affecting growth and development of crops. The existence grass not only compresses living space seedlings but also easily causes insect damage. Therefore, it is essential for weeding field. existing methods usually use manual or chemical herbicide spraying, which time-consuming laborious inefficient. With artificial intelligence modern agricultural technology, robots become effective means, attracted more attention...
Empowering models to dynamically accomplish tasks specified through natural language instructions represents a promising path toward more capable and general artificial intelligence. In this work, we introduce InstructSeq, an instruction-conditioned multi-modal modeling framework that unifies diverse vision flexible control handling of both visual textual data. InstructSeq employs multimodal transformer architecture encompassing visual, language, sequential modeling. We utilize encoder...