A5059192195- Domain Adaptation and Few-Shot Learning
- COVID-19 diagnosis using AI
- Advanced Image and Video Retrieval Techniques
- Optimization and Search Problems
- Image Processing Techniques and Applications
- Advanced Neural Network Applications
- Multimodal Machine Learning Applications
- Cancer-related molecular mechanisms research
- Robotics and Sensor-Based Localization
- Metaheuristic Optimization Algorithms Research
- Machine Learning and Data Classification
- Distributed Control Multi-Agent Systems
- Advanced Malware Detection Techniques
- Evolutionary Algorithms and Applications
- Reinforcement Learning in Robotics
- Robotic Path Planning Algorithms
- Internet Traffic Analysis and Secure E-voting
- Industrial Vision Systems and Defect Detection
- Information and Cyber Security
- Anomaly Detection Techniques and Applications
- Human Pose and Action Recognition
- Advanced Vision and Imaging
- Terrorism, Counterterrorism, and Political Violence
- Cybercrime and Law Enforcement Studies
- Hepatitis B Virus Studies
Lanzhou University of Technology
2019-2024
National Central University
2020-2023
Institute for Information Industry
2022
Analogic (United States)
2021
Few-shot learning for fine-grained image classification has gained recent attention in computer vision. Among the approaches few-shot learning, due to simplicity and effectiveness, metric-based methods are favorably state-of-the-art on many tasks. Most of assume a single similarity measure thus obtain feature space. However, if samples can simultaneously be well classified via two distinct measures, within class distribute more compactly smaller space, producing discriminative maps....
The main challenge for fine-grained few-shot image classification is to learn feature representations with higher inter-class and lower intra-class variations, a mere few labelled samples. Conventional learning methods however cannot be naively adopted this setting -- quick pilot study reveals that they in fact push the opposite (i.e., variations variations). To alleviate problem, prior works predominately use support set reconstruct query then utilize metric determine its category. Upon...
Few-shot fine-grained image classification has attracted considerable attention in recent years for its realistic setting to imitate how humans conduct recognition tasks. Metric-based few-shot classifiers have achieved high accuracies. However, their metric function usually requires two arguments of vectors, while transforming or reshaping three-dimensional feature maps vectors can result loss spatial information. Image reconstruction is thus involved retain more appearance details: the test...
The main challenge for fine-grained few-shot image classification is to learn feature representations with higher inter-class and lower intra-class variations, a mere few labelled samples. Conventional learning methods however cannot be naively adopted this setting - quick pilot study reveals that they in fact push the opposite (i.e., variations variations). To alleviate problem, prior works predominately use support set reconstruct query then utilize metric determine its category. Upon...
Due to the agile mobility of unmanned aerial vehicles (UAVs), UAVs become potential robotic platforms for search and rescue applications. Since making a decision (e.g., identify victims or termination mission) in missions is difficult, most systems rely on teleoperation. This research proposes novel telerobotic system consisting monitor, joystick eye tracker drone with RGBD camera. The experiments demonstrate that (1) human pilots can efficiently; (2) collected data rosbag format, which be...
Fine-grained image Classification is an important task in computer vision. The main challenge of the are that intra-class similarity large and training data points each class insufficient for a deep neural network. Intuitively, if we can learn more discriminative features detailed from fined-grained images, classification performance be improved. Considering channel attention features, spatial this paper proposes new mechanism by modifying Squeeze-and-Excitation block, mixed combining...
Finding an optimal search path is a NP-hard problem. Since one of human central activities, learning spatial behavior from operators way to solve problems. Utilizing the submodularity problems, this research proposes submodular inverse reinforcement (SIRL) algorithm learn humans' behavior. The experiments demonstrate that performance learned paths outperform state art approaches (e.g., MaxEnt IRL and DIRL).
Few-shot image classification aims to provide accurate predictions for novelty by learning from a limited number of samples. Classical few-shot methods usually use data augmentation and self-supervision compensate the lack training sample, introduce migration meta-learning pre-train model or accelerate optimization, which improves performance model. However, with small amount labeled sample data, these cannot meet requirements model's ability characterize features, resulting in that is...
The main challenge for fine-grained few-shot image classification is to learn feature representations with higher inter-class and lower intra-class variations, a mere few labelled samples. Conventional learning methods however cannot be naively adopted this setting -- quick pilot study reveals that they in fact push the opposite (i.e., variations variations). To alleviate problem, prior works predominately use support set reconstruct query then utilize metric determine its category. Upon...
With the increasing use of internet, cyber threats and malicious activities are becoming ubiquitous. To avoid unsuspecting attacks, gathering enough information about different is crucial. According to Pyramid Pain, Indicators Compromise (IOCs) simplest artifacts observe, which help security professionals design corresponding precautions. Cyber Threat Intelligence (CTI) data that presents current threat events, actors' targets, attack behaviors; hence, collecting analyzing CTI in advance can...
The purpose of the few-shot classification is to classify new categories, and each category contains few labeled samples. currently popular cross-domain uses a feature transformation layer transform features achieve enhancement, so as simulate various distributions in different domains during training process. However, due large differences distribution features, single cannot perform multiple transformations. To obtain change domains, diversified proposed based on original solve...