Image Analysis and Processing - ICIAP 2025

This two-volume set LNCS 16167-16168 constitutes the refereed proceedings of the 23rd International Conference on Image Analysis and Processing, ICIAP 2025, held in Rome, Italy, during September 15-19, 2025.

The 104 papers included in these proceedings were carefully reviewed and selected from 178 submissions. They covered a broad and evolving spectrum of topics in Computer Vision, Image Analysis, Pattern Recognition, and Machine Learning, spanning both foundational theories and impactful applications.

Incremental Object-Based Novelty Detection with Feedback Loop. - Zero-Shot Neural Architecture Search for Efficient Deep Stereo Matching. - DeepCS-TRD, a Deep Learning-based Cross-Section Tree Ring Detector. - KairosAD: A SAM-Based Model for Industrial Anomaly Detection on Embedded Devices. - SegDT: A Diffusion Transformer-Based Segmentation Model for Medical Imaging. - FS-SAM2: Adapting Segment Anything Model 2 for Few-Shot Semantic Segmentation via Low-Rank Adaptation. - Confidently Biased (ConB): A Per-Sample Confidence Approach for Unsupervised Model Debiasing. - A Reinforcement Learning Framework for Optimizing Predictive Maintenance of Photovoltaic Systems. - Unsupervised Segmentation for Reliable Estimation of Stratum Corneum Thickness Using LC-OCT. - MATE: Multimodal Agent that Talks and Empathizes. - Intrinsic Training Signals for Federated Learning Aggregation. - Disentangled Latent Spaces Facilitate Data-Driven Auxiliary Learning. - Deep Multimodal Fusion for 2D-3D Vineyard Scene Understanding. - There is No Model to Beat Them All: Recommendations for Deep Learning Model Selection when Training on Synthetic Images. - SVGauge: Towards Human-Aligned Evaluation for SVG Generation. - A Second-Order Perspective on Pruning at Initialization and Knowledge Transfer. - Depth and Event-Based Approaches for Human Detection and Pose Estimation. - Comparative Analysis of Lightweight Vision Models on Embedded Accelerator Devices. - Efficient Single Image Super-Resolution for Images with Spatially Varying Degradations. - MADPOT: Medical Anomaly Detection with CLIP Adaptation and Partial Optimal Transport. - Predicting Soccer Penalty Kick Direction Using Human Action Recognition. - AI-Driven Visual Monitoring of Industrial Assembly Tasks. - Benchmarking Content-Based Puzzle Solvers on Corrupted Jigsaw Puzzles. - Successful 3D Reconstructions of Underwater World War II Wreckage. - Explainable Machine Learning for Earthquakes: SHAP Interpretation of CNNs to Distinguish Seismic Spectrograms of Foreshocks and Aftershocks. - RoadFusion: Latent Diffusion Model for Pavement Defect Detection. - Historical Postcards Classification Combining Visual Content and Text Description. - VidCLearn: A Continual Learning Approach for Text-to-Video Generation. - Ensemble-Enhanced Semantic Segmentation for Efficient Waste Recycling. - FSMDet: Vision-Guided Feature Diffusion for Fully Sparse 3D Detector. - Efficient Odd-One Out Anomaly Detection. - Physics-Based 3D Simulation for Synthetic Data Generation and Failure Analysis in Packaging Stability Assessment. - Segmenting Anything in Architecture: a Tailored Approach to Segmenting Floor Plan Documents. - Automated Instance Label Generation for Traffic Domain to Enhance YOLACT-YOLO. - Accelerating YOLO with EOBranch: An Early Exit Approach for Adaptive Object Detection. - Mamba-OTR: a Mamba-based Solution for Online Take and Release Detection from Untrimmed Egocentric Video. - Zero-Shot Fibers Detection in Scanning Electron Micrographs. - 7Bench: a Comprehensive Benchmark for Layout-guided Text-to-image Models. - Efficient Attention on Microcontrollers via Linearized Transformer Blocks. - VS-Sim: A Synthetic Dataset for Viewpoint Shift Robustness. - Few-Shot Anomaly Classification by Learning to Inpaint Nominal Images. - AGCD-Net: Attention Guided Context Debiasing Network for Emotion Recognition. - Prompt-Based Consistent Video Colorization. - An Investigation on Incremental Learning from Unbalanced Streamed Data. - Improving Generative Data Augmentation with Prior-Knowledge for Dysplasia Grading of Colorectal Polyps. - Automatic Benchmarking of Large Multimodal Models via Iterative Experiment Programming. - High-Fidelity Infrared Scenario Simulation with Conditional CycleGAN. - Multi-Frame Alignment for Video Super-Resolution Using Attention. - D2R2: Deep-Learning Based Real-Time Dense Object Reconstruction. - Unsupervised Anomaly Localization in the Wild via Token Optimization and Test-Time Score Correction. - Enhancing Scene Segmentation Using Spatial Relationships: A Case Study on Byzantine Seals. - Image Specific Protection Against Manipulation.