A5101837320- Smart Agriculture and AI
- Spectroscopy and Chemometric Analyses
- Remote Sensing in Agriculture
- Domain Adaptation and Few-Shot Learning
- Industrial Vision Systems and Defect Detection
- Multimodal Machine Learning Applications
- Remote Sensing and Land Use
- Human Pose and Action Recognition
- Advanced Neural Network Applications
- Advanced Image and Video Retrieval Techniques
- Ergonomics and Musculoskeletal Disorders
- Gait Recognition and Analysis
- Identification and Quantification in Food
- Color perception and design
- Machine Learning and ELM
- COVID-19 diagnosis using AI
- Horticultural and Viticultural Research
- Visual Attention and Saliency Detection
- Anomaly Detection Techniques and Applications
- Hearing Impairment and Communication
- Spatial Cognition and Navigation
- Video Analysis and Summarization
- Advanced Optical Network Technologies
- Image and Video Stabilization
- Software System Performance and Reliability
Shandong Agricultural University
2021-2024
University at Buffalo, State University of New York
2020-2023
Hisense (China)
2021
Nanyang Technological University
2018
Huazhong University of Science and Technology
2015-2017
Soochow University
2007
We propose SonicASL, a real-time gesture recognition system that can recognize sign language gestures on the fly, leveraging front-facing microphones and speakers added to commodity earphones worn by someone facing person making gestures. In user study (N=8), we evaluate performance of various at both word sentence levels. Given 42 frequently used individual words 30 meaningful sentences, SonicASL achieve an accuracy 93.8% 90.6% for word-level sentence-level recognition, respectively. The...
Crop cultivar identification is an important aspect in agricultural systems. Traditional solutions involve excessive human interventions, which labor-intensive and timeconsuming. In addition, a typical task of fine-grained visual categorization (FGVC). Compared with other common topics FGVC, studies this problem are somewhat lagging limited. paper, targeting four Chinese maize cultivars Jundan No.20, Wuyue No.3, Nongda No.108, Zhengdan No.958, we first consider the identifying based on its...
This study investigates the low mechanization level of grape picking, and problems associated with difficult location picking points in three-dimensional space. A method for rapidly locating table grapes based on an infrared tube was proposed this paper. Firstly, Otsu algorithm maximum connected domain were used to obtain image target grape, which realized fast recognition segmentation two-dimensional Secondly, a device grape-picking designed, resolved technical problem related difficulty...
Neural architecture search (NAS) proves to be among the best approaches for many tasks by generating an application-adaptive neural architectures, which are still challenged high computational cost and memory consumption. At same time, 1-bit convolutional networks (CNNs) with binarized weights activations show their potential resource-limited embedded devices. One natural approach is use CNNs reduce computation of NAS taking advantage strengths each in a unified framework. To this end,...
Crop segmentation is a frequently concerned problem for computer vision applications in agriculture. Tassel typical agronomic trait the crop breeding process. characterization also requires fine-grained shape extraction. However, previous methods are usually dependent of category, which hard to transfer other cultivars with different colors. To address this, goal this study develop feasible method that can deal categories simultaneously and easy transfer. Targeted on maize, we proposed...
Continual Learning (CL) has achieved rapid progress in recent years. However, it is still largely unknown how to determine whether a CL model trustworthy and foster its trustworthiness. This work focuses on evaluating improving the robustness corruptions of existing models. Our empirical evaluation results show that state-of-the-art (SOTA) models are particularly vulnerable various data during testing. To make them robust deployed safety-critical scenarios, we propose meta-learning framework...
Cultivar identification is an important aspect in agriculture and also a typical task of fine-grained visual categorization (FGVC). In comparison with other common topics FGVC, studies on this problem are somewhat lagged limited. paper, targeting four Chinese maize cultivars Jundan No.20, Wuyue No.3, Nongda No.108, Zhengdan No.958, we first consider the identifying cultivar based its tassel characteristics. Technically, effective convolutional neural network (CNN) feature encoding pipeline...
Given a video clip that contains only one type of action (e.g., golfing), the goal recognition is to recognize this category from given set types. To deliver fast response for practical applications, existing works have been endevouring on processing leading frames input video. In our view, informative key extracted `partial video' should be used performing task. This will not further speed up process due less amount data processed but also achieve higher accuracy owing more distinctive...
Forecasting human actions and motion trajectories address the problem of predicting what a person is going to do next how they will perform it. This crucial in wide range applications, such as assisted living future co-robotic settings. We propose simultaneously learn action-related dynamics while existing works them independently. paper presents method jointly forecast categories action skeletal joint pose, allowing two tasks reinforce each other. As result, our system can predict that...
Aiming at the problem of low intelligence in automatic navigation cuttage and film covering multi-functional machine for tunnels, this study proposed a line extraction method based on improved YOLOv5s model, which can achieve accurate lines two planting methods seedling transplanting direct seeding. Firstly, we pre-processed acquired images using inverse perspective transformation. Next, Coordinate Attention Ghost modules were applied to improve architecture, increasing detection accuracy...