Robotics, Computer Vision and Intelligent Systems. 4th International Conference, ROBOVIS 2024 Rome, Italy, February 25–27, 2024 Proceedings

Preface
Organization
Invited Speakers
	Learning with Privileged Information and Distillation for Multimodal Video Classification
	There’s Plenty of Room at the Bottom: Opportunites and Challenges for Microrobotics
	Physical Models and Machine Learning for Photography and Astronomy
	Contents
Compute Optimal Waiting Times for Collaborative Route Planning
	1 Introduction
	2 Related Work
	3 Optimal Waiting Times
		3.1 Introductory Considerations
		3.2 Model and Definitions
		3.3 The Idea
		3.4 The Algorithm
	4 Extending the Approach for Collaborative Routing
		4.1 Introduction
		4.2 Critical Collisions
		4.3 Resolving Critical Collisions
		4.4 Putting All Together
	5 Evaluation
	6 Conclusions
	References
Robot Vision and Deep Learning for Automated Planogram Compliance in Retail
	1 Introduction
	2 Images Capturing and Shelves Image Reconstruction
		2.1 Robot Vision System and Images Capturing
		2.2 Shelves Image Reconstruction
	3 Planogram and Database
	4 Object Detection
	5 Planogram Compliance
	6 Experimental Results
	7 Conclusions
	References
Park Marking Detection and Tracking Based on a Vehicle On-Board System of Fisheye Cameras
	1 Introduction
	2 Related Work
	3 Methodology
		3.1 Surround View System
		3.2 Park Marking Detection
		3.3 3D Ground Projection
		3.4 Tracking and Multi-camera Fusion
	4 Experiments
		4.1 Dataset
		4.2 Quantitative Results
		4.3 Qualitative Results
	5 Conclusions
	References
Analysis of Age Invariant Face Recognition Efficiency Using Face Feature Vectors
	1 Introduction
	2 Previous Work
	3 Face Recognition Methods
		3.1 InsightFace
	4 Datasets
		4.1 AgeDB
		4.2 CASIA
		4.3 LFW
	5 Measurements Used in the Experiments
		5.1 Similarity Between FFVs
		5.2 Mean Average Precision
		5.3 Bhattacharyya Measurements
	6 Results
		6.1 Similarity Distributions
		6.2 Pipeline Performance
		6.3 Age Group Performance
		6.4 Correlation Between Measurements
	7 Discussion
	8 Conclusion
		8.1 Future Work
	References
Uncertainty Driven Active Learning for Image Segmentation in Underwater Inspection
	1 Introduction
	2 Related Work
	3 Method
	4 Experimental Evaluation
	5 Conclusion
	References
Enhancing Connected Cooperative ADAS: Deep Learning Perception in an Embedded System Utilizing Fisheye Cameras
	1 Introduction
	2 Related Work
	3 System Architecture
	4 Perception
		4.1 Embedded Computing
	5 V2X Communication
	6 LDM
	7 Applications
	8 Experiments and Results
		8.1 Experimental Setup
		8.2 Perception Evaluation
		8.3 Communication Evaluation
		8.4 System Limits
	9 Conclusions
	References
Weapon Detection Using PTZ Cameras
	1 Introduction
	2 Related Work
	3 Tracking/Zoom System
	4 Experiments
		4.1 Experimental Setup
		4.2 Model Training
		4.3 Results
	5 Conclusions and Future Work
	References
Improving Semantic Mapping with Prior Object Dimensions Extracted from 3D Models
	1 Introduction
	2 Method Description
		2.1 The Overall Semantic Mapping Process
		2.2 Prior Knowledge
		2.3 2D Geometric Association Method
		2.4 Algorithm Description
	3 Experimentation
		3.1 Experimental Setup
		3.2 Tuning the Scoring Function Weights
		3.3 Evaluation of Polygon Simplification and Foreground Edges Selection
		3.4 Evaluation of the Association Algorithm
	4 Conclusion and Perspectives
	References
Offline Deep Model Predictive Control (MPC) for Visual Navigation
	1 Introduction
	2 Related Work
	3 Proposed Approach
		3.1 Simulation Environment
		3.2 Future Image Prediction, ViewNet
		3.3 Control Policy with VelocityNet
		3.4 Training VelocityNet
		3.5 Visual Trajectory Following with VelocityNet, Control Algorithm
	4 Experimental Setup
		4.1 Network Structure
		4.2 Training
		4.3 Parameters
	5 Results Analysis
		5.1 Linear Translation
		5.2 Pure Rotation
		5.3 Combined Translation and Rotation
		5.4 Statistical Analysis
	6 Conclusion
	References
BiGSiD: Bionic Grasping with Edge-AI Slip Detection
	1 Introduction
	2 Related Work
	3 Data Collection
		3.1 Hardware Setting
		3.2 Dataset
	4 Proposed Method and Training the Model
		4.1 Proposed Method
		4.2 Training the Model
		4.3 Experiments and Results
	5 AI Edge-Device
		5.1 Bringing AI to the Edge-Device
	6 Conclusion
	References
GAT-POSE: Graph Autoencoder-Transformer Fusion for Future Pose Prediction
	1 Introduction
	2 Related Works
		2.1 Probabilistic Prediction
		2.2 Deterministic Prediction
	3 Methodology
		3.1 Problem Formulation
		3.2 Tokenization
		3.3 Transformer Autoencoder with Future Masking
	4 Experiment and Evaluation
		4.1 Human3.6M
		4.2 Evaluation Metrics
		4.3 Results and Comparative Analysis
	5 Conclusions and Future Work
	References
UCorr: Wire Detection and Depth Estimation for Autonomous Drones
	1 Introduction
	2 Related Work
		2.1 Wire Detection
		2.2 Depth Estimation
	3 Method
		3.1 Motivation
		3.2 UCorr Network Architecture
		3.3 Loss Function
	4 Experiments
		4.1 Data
		4.2 Metrics
		4.3 Training
		4.4 Results
		4.5 Ablation Studies
	5 Discussion
	6 Conclusion
	References
A Quality-Based Criteria for Efficient View Selection
	1 Introduction
	2 Next-Best-View Selection
	3 Selected NBV Algorithm
		3.1 A Global Max-Flow Based Method
		3.2 Problem Adaptation for Accuracy Improvement
	4 Estimation of 3D Model Quality
		4.1 From Existing 3D Quality Metrics...
		4.2 ...to a Quality Metric for NBV
	5 Experiments on Simulated Data
		5.1 Experimental Setup
		5.2 Results and Analysis
	6 Experiments on Synthetic Images
		6.1 Experimental Setup
		6.2 Results and Analysis
	7 Conclusion
	References
Multi-UAV Weed Spraying
	1 Introduction
	2 Problem Definition
	3 Proposed System
		3.1 Clustering
		3.2 Finding the Proper Way-Points
		3.3 Path-Planning
		3.4 Online Collision Avoidance
	4 Mapping
	5 Spraying Technical Information
	6 Experimentation
	7 Conclusion and Future Work
	References
Human Comfort Factors in People Navigation: Literature Review, Taxonomy and Framework
	1 Introduction
	2 People Navigation: Definition, Review and Classification
		2.1 Definitions
		2.2 Related Systematic Review
		2.3 Taxonomy
	3 Human Comfort Factors in People Navigation
		3.1 Human Comfort Factors in Navigation Systems
		3.2 Summary of Human Comfort Factors in People Navigation
	4 Towards a People Navigation Framework
		4.1 Human-Aware Navigation
		4.2 Implementing Passenger Comfort Factors
	5 Conclusion
	References
Region Prediction for Efficient Robot Localization on Large Maps
	1 Introduction
	2 Background and Related Works
	3 Methodology
		3.1 Map Clustering
		3.2 Region Prediction
		3.3 Integration in RTAB-Map
	4 Experiments and Results
		4.1 Prediction Accuracy
		4.2 Loop Closure Detection
		4.3 Pose Estimation Accuracy
	5 Conclusions and Future Works
	References
Utilizing Dataset Affinity Prediction in Object Detection to Assess Training Data
	1 Introduction and Related Work
	2 Dataset Affinity Prediction
	3 Dataset Alignment
	4 Evaluation
	5 Conclusion
	References
Optimizing Mobile Robot Navigation Through Neuro-Symbolic Fusion of Deep Deterministic Policy Gradient (DDPG) and Fuzzy Logic
	1 Introduction
		1.1 Deep Deterministic Policy Gradient Algorithm
		1.2 Neuro-Symbolic Approach Through Fuzzy Logic in Reward Function
	2 Methods
		2.1 Overview of the Hybrid System Architecture of DDPG and Fuzzy Logic
		2.2 Constructing DDPG Agent
		2.3 Constructing Reward Function Using Fuzzy Logic
		2.4 Agents’ Training
	3 Experiments and Results
		3.1 DDPG Agent Training Results
		3.2 Mobile Robot Simulation Results
	4 Discussions
		4.1 Overall Reward for DDPG Training
		4.2 Mobile Robot Simulation Performance
		4.3 Implementation of Fuzzy Logic in Reward Function
	5 Conclusion
	References
DAFDeTr: Deformable Attention Fusion Based 3D Detection Transformer
	1 Introduction
	2 Related Work
		2.1 LiDAR-Only Approaches
		2.2 Multi-modal Approaches
	3 Methodology
		3.1 Backbones
		3.2 LiDAR Encoder
		3.3 Query Initialization
		3.4 LiDAR Decoder
		3.5 Image Decoder
		3.6 Loss
	4 Experiments
		4.1 Implementation Details
		4.2 Results and Discussion
		4.3 Ablation Studies
		4.4 Qualitative Results
	5 Conclusion
	References
MDC-Net: Multimodal Detection and Captioning Network for Steel Surface Defects
	1 Introduction
	2 Literature Review
	3 Methodology
		3.1 Dataset Description
		3.2 Dataset Preprocessing
		3.3 Model Architecture
	4 Implementation and Training
		4.1 Hyperparameter Optimization
		4.2 Reproducibility and Model Consistency
		4.3 Loss Function and Optimization
	5 Results and Discussion
	6 Conclusion and Future Direction
	References
Operational Modeling of Temporal Intervals for Intelligent Systems
	1 Introduction
	2 Proposed Approach
	3 Application and Discussion
	4 Conclusions
	References
A Meta-MDP Approach for Information Gathering Heterogeneous Multi-agent Systems
	1 Introduction
	2 Background
		2.1 Markov Decision Process
		2.2 Partially Observable Markov Decision Process
	3 Model
		3.1 The Exploration POMDPs
		3.2 The Meta-MDP
		3.3 Approximating the Meta Transition Function
		3.4 Complexity Analysis
	4 Experimentation
		4.1 Experimental Setup
		4.2 Comparative Models and Measures
		4.3 Experimental and Theoretical Results
	5 Conclusion
	6 Future Work
	References
Interacting with a Visuotactile Countertop
	1 Introduction
	2 Previous Work
		2.1 Dense Tactile Sensors
		2.2 Visual Tactile Displays
	3 The STS Display
		3.1 Identifying Interaction Objects
		3.2 Tactile/Visual Duty Cycle
		3.3 The STS-D Display
	4 The Interactive Countertop Scenario
		4.1 Sample Interaction with the STS-d
		4.2 Interaction Object Dataset
		4.3 Training and Recognition
		4.4 Estimating Object Pose and Trajectory
	5 Summary and Ongoing Work
	References
A Color Event-Based Camera Emulator for Robot Vision
	1 Introduction
	2 Related Work
		2.1 Event-Based Cameras Working Principle
		2.2 Simulators and Emulators
		2.3 Hardware Implementation of Color Event-Based Cameras
	3 Color Event Camera Emulator for Robotic Vision
		3.1 ROS Software Integration
	4 Qualitative Results
		4.1 Datasets
		4.2 Qualitative Analysis
		4.3 Example Application
		4.4 Processing Times
	5 Software Implementation Discussion
	6 Conclusions and Future Work
	References
Fast Point Cloud to Mesh Reconstruction for Deformable Object Tracking
	1 Introduction
	2 Related Work
		2.1 3D Mesh Reconstruction of a Point Cloud
		2.2 4D Mesh Reconstruction of Point Clouds
	3 Methodology
		3.1 Auto-Encoder
		3.2 Real-NVP
		3.3 Network Architecture
		3.4 Inference
		3.5 Dataset
	4 Experiments
		4.1 Dataset Generation
		4.2 Discussion
	5 Conclusion and Future Work
	References
Estimation of Optimal Gripper Configuration Through an Embedded Array of Proximity Sensors
	1 Introduction
	2 Related Works
	3 The Hybrid-Active Gripper
	4 Embedded Array of Proximity Sensors
	5 Estimation of Grasp Pose Configuration
	6 Analysis of Gripper Configuration Estimation
	7 Results Analysis of Grasp Pose Estimation
		7.1 Pose Estimation Gripper with Two Fingers
		7.2 Pose Estimation Gripper with Three Fingers
	8 Conclusion
	References
The Twinning Technique of the SyncLMKD Method
	1 Introduction
	2 The Challenges for Twinning and the Solutions
		2.1 Proposed Approach for Twinning Technique
		2.2 Control
		2.3 Timestamps
		2.4 Mirroring
		2.5 Predictability and Effect Function
	3 Evaluation
	4 Conclusion
	References
Intuitive Multi-modal Human-Robot Interaction via Posture and Voice
	1 Introduction
	2 Related Work
	3 Multi-modal Human-Robot-Interaction
		3.1 Problem Formulation
		3.2 Materials and Methods
		3.3 Experimental Setup
		3.4 Experimental Results
	4 Chatbot
	5 Conclusion
	References
Virtual Model of a Robotic Arm Digital Twin with MuJoCo
	1 Introduction
		1.1 Related Works
		1.2 Contribution
	2 Hardware and Software Architecture
		2.1 Physical Robot
		2.2 Physical Communication
		2.3 Digital Entity
	3 Proposed Frameworks
		3.1 Digital Shadow
		3.2 Digital Driven
		3.3 Digital Twin
	4 Preliminary Tests
	5 Discussion
	6 Concluding Remarks and Futures Works
	References
Author Index