A5025644552- Robot Manipulation and Learning
- Advanced SAR Imaging Techniques
- Advanced Vision and Imaging
- Synthetic Aperture Radar (SAR) Applications and Techniques
- Industrial Vision Systems and Defect Detection
- Soil Moisture and Remote Sensing
- Image Enhancement Techniques
- Surface Roughness and Optical Measurements
- Advanced Measurement and Metrology Techniques
- AI in cancer detection
- Advanced Neural Network Applications
- COVID-19 diagnosis using AI
- Advanced Measurement and Detection Methods
- Visual Attention and Saliency Detection
- Soft Robotics and Applications
- Sparse and Compressive Sensing Techniques
- Face and Expression Recognition
- Healthcare and Venom Research
- Blind Source Separation Techniques
- Welding Techniques and Residual Stresses
- Digital Imaging for Blood Diseases
- Robotics and Sensor-Based Localization
- Fatigue and fracture mechanics
- Image and Signal Denoising Methods
- Computer Graphics and Visualization Techniques
Wuhan University of Technology
2025
Tsinghua University
2022-2025
Transparent and reflective objects are omnipresent in our daily life, but their unique visual optical characteristics notoriously challenging even for state-of-the-art deep networks of semantic segmentation. To alleviate this challenge, we construct a new large-scale real-world RGB-D dataset called TROSD, which is more comprehensive than existing datasets transparent object Our TROSD contains 11,060 images with three classes terms objects, others, covering variety scenes. Together the...
Traditional affordance segmentation on 3D point cloud objects requires massive amounts of annotated training data and can only make predictions within predefined classes tasks. To overcome these limitations, we propose a variation-robust few-shot network (VRNet) for robotic manipulation, which several annotations novel object manipulation In particular, design an orientation-tolerant feature extractor to address pose variation between support query objects, present multi-scale label...
This paper presents an Enhanced Multilinear Principal Component Analysis (EMPCA) algorithm, improved variant of the traditional (MPCA) tailored for efficient dimensionality reduction in high-dimensional data, particularly image analysis tasks. EMPCA integrates random singular value decomposition to reduce computational complexity while maintaining data integrity. Additionally, it innovatively combines method with Mask R-CNN enhancing accuracy segmentation. Leveraging tensors, achieves that...
Abstract In order to solve the problem of unclear relationship between aerodynamic characteristics aircraft engine fan blade airfoil (such as leading edge radius and wedge angle) its resistance foreign body damage (FOD) sensitivity analysis under data redundancy conditions that cannot be achieved by traditional methods, a method based on K-means clustering was proposed. First, simulated impact tests were conducted TC4 titanium alloy specimens using an air cannon. The step-loading test used...
Transparent and reflective objects, which are common in our everyday lives, present a significant challenge to 3D imaging techniques due their unique visual optical properties. Faced with these types of RGB-D cameras fail capture the real depth value accurate spatial information. To address this issue, we propose DITR, diffusion-based Depth Inpainting framework specifically designed for Reflective objects. This network consists two stages, including Region Proposal stage stage. DITR...
Manipulating unseen articulated objects through visual feedback is a critical but challenging task for real robots. Existing learning-based solutions mainly focus on affordance learning or other pre-trained models to guide manipulation policies, which face challenges novel instances in real-world scenarios. In this letter, we propose part-guided 3D RL framework, can learn manipulate without demonstrations. We combine the strengths of 2D segmentation and improve efficiency policy training. To...
Multi-band polarimetric synthetic aperture radar (PolSAR) has significant advantage in information extraction. However, the demanding acquisition requirement greatly prohibits its development. Typically, compared to low-frequency band PolSAR data, high-frequency suffers more severe data insufficiency. In this paper, authors proposed resolve issue by simulating Ka-band images from X-band images. For purpose, a conditional Generative Adversial Network (cGAN) based X-to-Ka image transfer...
Leukemia is a common, multiple and dangerous blood disease, whose early diagnosis treatment are very important. At present, the of leukemia heavily relies on morphological examination cell images by pathologists, which tedious time-consuming. Meanwhile, diagnostic results highly subjective, may lead to misdiagnosis missed diagnosis. To address gap above, we proposed an improved Vision Transformer model for recognition. First, faster R-CNN network was used locate extract individual slices...
Manipulating unseen articulated objects through visual feedback is a critical but challenging task for real robots. Existing learning-based solutions mainly focus on affordance learning or other pre-trained models to guide manipulation policies, which face challenges novel instances in real-world scenarios. In this paper, we propose part-guided 3D RL framework, can learn manipulate without demonstrations. We combine the strengths of 2D segmentation and improve efficiency policy training. To...
Industrial anomaly detection (IAD) plays a crucial role in the maintenance and quality control of manufacturing processes. In this paper, we propose novel approach, Vision-Language Anomaly Detection via Contrastive Cross-Modal Training (CLAD), which leverages large vision-language models (LVLMs) to improve both localization industrial settings. CLAD aligns visual textual features into shared embedding space using contrastive learning, ensuring that normal instances are grouped together while...
Data insufficiency poses a significant challenge in Ka-band Polarimetric Synthetic Aperture Radar (PolSAR) applications. Traditional PolSAR simulation approaches fail to conquer this issue due the intricate modeling and computational complexities induced by high-frequency. In paper, authors propose mitigate through neural style transfer. An X2Ka translation network is proposed transfer X-band images Ka-band. Leveraging well-verified generative Pix2Pix, adapt it accommodate specific...