A5008655688- Robot Manipulation and Learning
- Advanced Image and Video Retrieval Techniques
- Multimodal Machine Learning Applications
- Domain Adaptation and Few-Shot Learning
- Robotics and Sensor-Based Localization
- Advanced Neural Network Applications
- Human Pose and Action Recognition
- Robotic Path Planning Algorithms
- Robotics and Automated Systems
- Reinforcement Learning in Robotics
- Image Retrieval and Classification Techniques
- Modular Robots and Swarm Intelligence
- Robotic Locomotion and Control
- Hand Gesture Recognition Systems
- Visual Attention and Saliency Detection
- Advanced Vision and Imaging
- Soft Robotics and Applications
- Semantic Web and Ontologies
- AI-based Problem Solving and Planning
- Human Motion and Animation
- Generative Adversarial Networks and Image Synthesis
- Industrial Vision Systems and Defect Detection
- Berry genetics and cultivation research
- Distributed Control Multi-Agent Systems
- Genetic Neurodegenerative Diseases
University of Groningen
2018-2024
University of Aveiro
2014-2019
Islamic Azad University, Isfahan
2009-2016
Instituto de Telecomunicações
2016
Islamic Azad University, Tehran
2010
Recovering object pose in a crowd is challenging task due to severe occlusions and clutters. In active scenario, whenever an observer fails recover the poses of objects from current view point, able determine next position captures new scene another point improve knowledge environment, which may reduce 6D estimation uncertainty. We propose complete multi-view framework recognize 6DOF multiple instances crowded scene. include several components vision setting increase accuracy: Hypothesis...
Visual target navigation in unknown environments is a crucial problem robotics. Despite extensive investigation of classical and learning-based approaches the past, robots lack common-sense knowledge about household objects layouts. Prior state-of-the-art to this task rely on learning priors during training typically require significant expensive resources time for learning. To address this, we propose new framework visual that leverages Large Language Models (LLM) impart common sense object...
matching. In this system main stage is the isolation of license plate, from digital image car obtained by a camera under different circumstances such as illumination, slop, distance, and angle. The algorithm starts with preprocessing signal conditioning. Next plate localized using morphological operators. Then template matching scheme will be used to recognize digits characters within plate. tested on Iranian images, performance was 97.3% correct plates identification localization 92%...
Nowadays robots play an increasingly important role in our daily life. In human-centered environments, often encounter piles of objects, packed items, or isolated objects. Therefore, a robot must be able to grasp and manipulate different objects various situations help humans with tasks. this paper, we propose multi-view deep learning approach handle robust object grasping human-centric domains. particular, takes point cloud arbitrary as input, then, generates orthographic views the given...
Nowadays, service robots are appearing more and in our daily life. For this type of robot, open-ended object category learning recognition is necessary since no matter how extensive the training data used for batch learning, robot might be faced with a <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">new object</i> when operating real-world environment. In article, we present xmlns:xlink="http://www.w3.org/1999/xlink">OrthographicNet</i> ,...
Abstract Simultaneous object recognition and pose estimation are two key functionalities for robots to safely interact with humans as well environments. Although both use visual input, most state of the art tackles them separate problems since former needs a view-invariant representation, while necessitates view-dependent description. Nowadays, multi-view convolutional neural network (MVCNN) approaches show state-of-the-art classification performance. MVCNN has been widely explored, there...
Abstract To aid humans in everyday tasks, robots need to know which objects exist the scene, where they are, and how grasp manipulate them different situations. Therefore, object recognition grasping are two key functionalities for autonomous robots. Most state-of-the-art approaches treat as separate problems, even though both use visual input. Furthermore, knowledge of robot is fixed after training phase. In such cases, if encounters new categories, it must be retrained incorporate...
This paper addresses the problem of grounding semantic representations in intelligent service robots. In particular, this work contributes to addressing two important aspects, namely anchoring object symbols into perception objects and category known instances categories. The discusses memory requirements for storing both perceptual data and, based on analysis these requirements, proposes an approach components, a memory. perception, memory, learning interaction capabilities, are main focus...
Robots are still not able to grasp all unforeseen objects. Finding a proper configuration, i.e. the position and orientation of arm relative object, is challenging. One approach for grasping objects recognize an appropriate configuration from previous demonstrations. The underlying assumption in this that new similar known ones (i.e. they familiar) can be grasped way. However finding representation similarity metric main challenge developing familiar In paper, interactive object view...
This work focuses on the problem of visual target navigation, which is very important for autonomous robots as it closely related to high-level tasks. To find a special object in unknown environments, classical and learning-based approaches are fundamental components navigation that have been investigated thoroughly past. However, due difficulty representation complicated scenes learning policy, previous methods still not adequate, especially large scenes. Hence, we propose novel framework...
Robots operating in human-centered environments, such as retail stores, restaurants, and households, are often required to distinguish between similar objects different contexts with a high degree of accuracy. However, fine-grained object recognition remains challenge robotics due the intra-category low inter-category dissimilarities. In addition, limited number 3D datasets poses significant problem addressing this issue effectively. paper, we propose hybrid multi-modal Vision Transformer...
Service robots are increasingly integrating into our daily lives to help us with various tasks. In such environments, frequently face new objects while working in the environment and need learn them an open-ended fashion. Furthermore, must be able recognize a wide range of object categories. this paper, we present lifelong ensemble learning approach based on multiple representations address few-shot recognition problem. particular, form methods deep handcrafted 3D shape descriptors. To...
Intelligent service robots should be able to improve their knowledge from accumulated experiences through continuous interaction with the environment, and in particular humans. A human user may guide process of experience acquisition, teaching new concepts, or correcting insufficient erroneous concepts interaction. This paper reports on work towards interactive learning objects robot activities an incremental open-ended way. In particular, this addresses human-robot gathering. The robot's...
In open-ended domains, robots must continuously learn new object categories. When the training sets are created offline, it is not possible to ensure their representativeness with respect categories and features system will find when operating online. Bag of Words model, visual codebooks usually constructed from offline. This might lead non-discriminative words and, as a consequence, poor recognition performance. paper proposes which concurrently learns in an incremental online fashion both...
Abstract Despite the recent success of state-of-the-art 3D object recognition approaches, service robots still frequently fail to recognize many objects in real human-centric environments. For these robots, is a challenging task due high demand for accurate and real-time response under changing unpredictable environmental conditions. Most approaches use either shape information only ignore role color or vice versa. Furthermore, they mainly utilize $$L_n$$ <mml:math...
Humans learn to recognize and manipulate new objects in lifelong settings without forgetting the previously gained knowledge under non-stationary sequential conditions. In autonomous systems, agents also need mitigate similar behaviour continually object categories adapt environments. most conventional deep neural networks, this is not possible due problem of catastrophic forgetting, where newly overwrites existing representations. Furthermore, state-of-the-art models excel either...
The Vision Transformer (ViT) architecture has established its place in computer vision literature, however, training ViTs for RGB-D object recognition remains an understudied topic, viewed recent literature only through the lens of multi-task pretraining multiple modalities. Such approaches are often computationally intensive, relying on scale datasets to align RGB with 3D information. In this work, we propose a simple yet strong recipe transferring pretrained domains recognition, focusing...
Service robots are expected to be more autonomous and work effectively in human-centric environments. This implies that should have special capabilities, such as learning from past experiences real-time object category recognition. paper proposes an open-ended 3D recognition system which concurrently learns both the categories statistical features for encoding objects. In particular, we propose extension of Latent Dirichlet Allocation learn structural semantic (i.e., visual topics),...
Service robots are expected to autonomously and efficiently work in human-centric environments. For this type of robots, object perception manipulation challenging tasks due need for accurate real-time response. This paper presents an interactive open-ended learning approach recognize multiple objects their grasp affordances concurrently. is important contribution the field service since no matter how extensive training data used batch learning, a robot might always be confronted with...