SPECIAL SESSION

Special Session Ⅰ: Algorithm Optimization and Applications Based on Machine Learning and Deep Learning

Session Chairs: Assoc. Prof. Lifeng Yin, Dalian Jiaotong University, China

Assoc. Prof. Miao Wang, Henan University of Engineering, China

Keywords: Supervised Learning; Unsupervised Learning; Reinforcement Learning; Deep Learning

Information: Algorithm optimization and applications based on machine learning and deep learning are important research directions in the field of modern artificial intelligence, aimed at enhancing model performance and computational efficiency. Optimization methods include hyperparameter tuning, feature selection, training process optimization, and model compression, all designed to improve the accuracy and responsiveness of algorithms. These optimization techniques are widely applied in areas such as natural language processing, computer vision, healthcare, and fintech, contributing to higher predictive accuracy and real-time decision-making capabilities.

Below is an incomplete list of potential topics to be covered in the Special Session:

  • Optimization and Applications of Supervised Learning Algorithms  

  • Optimization and Applications of Unsupervised Learning Algorithms  

  • Optimization and Applications of Reinforcement Learning Algorithms  

  • Optimization and Applications Based on YOLOv8 Algorithm  

  • Optimization and Applications Based on Graph Neural Networks  

  • Research on the Application of Deep Learning in Multimodal Data Fusion

Submission Deadline: January 10, 2025


Special Session Ⅱ: Design and Training Strategies for Deep Learning Models Under Extreme Data Conditions and Computational Constraints

Session Chair: Assoc. Prof. Shuwei Huo, Northeastern University, China

Keywords: Extreme Data Learning; Few-shot Learning; Low-quality Data Learning; Weakly/Unsupervised Learning; Lightweight Model Design; Efficient Training Strategies

Information: Recently, general-purpose foundation models have achieved breakthrough progress in artificial intelligence, demonstrating exceptional performance and broad application prospects. However, the practical deployment of such models faces dual challenges in data resources and computational capabilities. On the data front, these models impose stringent requirements on both the scale and quality of training corpora, demanding not only massive high-quality data support but also professional manual annotations. This high-cost data acquisition mechanism significantly constrains the models' applicability in specific domains. On the computational front, both training and inference processes require robust computational infrastructure, and this resource-intensive characteristic substantially limits their practical implementation in real-world scenarios. To overcome these technical bottlenecks and promote innovation and development of intelligent models under resource-constrained conditions, this special issue cordially invites experts and scholars from both academia and industry to engage in in-depth discussions on intelligent model architecture design and training optimization methods under extreme data and computational constraints.

Scope of Submissions includes but is not limited to:

  • Few-shot/Zero-shot Deep Learning Algorithms

  • Deep Learning Methods for Low-quality Data Scenarios

  • Deep Learning Approaches for Weakly/Unlabeled Data

  • Lightweight Deep Architecture Design

  • Efficient Training Strategies under Resource-Constrained Environments

Submission Deadline: March 31, 2025


Special Session Ⅲ: Frontiers of Medical Artificial Intelligence

Session Chair: Assit. Prof. Wenlong Ming, Nanjing University of Information Science and Technology, China

Keywords: Medical Image Analysis; Bioinformatics; Foundation Model; Multi-modality & multi-omics; Precision Medicine

Information: With the development of AI and foundation models, the efficient integration of multimodal data—such as medical imaging, omics, and electronic health records—is driving rapid advancements in precision medicine. The interaction of multimodal data and the application of large models open new possibilities for early disease diagnosis, treatment optimization, and personalized healthcare. This topic will focus on how to leverage foundation model technology, image computing, bioinformatics, and multi-omics data to uncover latent relationships within multi-source data, advancing scientific understanding in disease diagnosis, prognosis prediction, and personalized therapy. It will also address key technical challenges in integrating bioinformatics and multi-omics data, showcasing innovative research cases with biological interpretability and clinical relevance. Through interdisciplinary exchange and collaboration, this conference aims to advance foundational research and clinical applications within the field of precision medicine.

Scope of Submissions includes but is not limited to:

  • Medical Image Segmentation

  • Medical Imaging Analysis

  • Radiomics

  • Computational Pathology

  • Medical Natural Language Processing

  • Medical Foundation Model

  • Multi-modality Data Fusion

  • Omics Data Analysis

  • Multi-omics Data Fusion

  • Intelligent Bioinformatics

Submission Deadline: March 23, 2025


Special Session Ⅳ: Scene Perception and Reconstruction for Autonomous Driving

Session Chair: Assoc. Prof. Yuanzhouhan Cao, Beijing Jiaotong University, China

Keywords: Autonomous Driving; Multi-source Data Fusion;Dynamic Scene Perception; 3D Scene Reconstruction

Information: Autonomous driving is at the core of intelligent transportation development. It not only has the potential to alleviate traffic congestion, reduce accidents, and improve travel efficiency, but also represents a critical step toward building smart cities and achieving sustainable transportation. Currently, autonomous driving technology is rapidly advancing toward higher levels of automation; however, numerous challenges still stand in the way of fully autonomous driving. Particularly in complex, dynamic environments, achieving high-precision perception, real-time responsiveness, and environment reconstruction requires overcoming significant technical hurdles such as efficient multi-source data fusion, real-time dynamic perception, and addressing environmental uncertainty.


This session will cover a range of key technologies, from multi-sensor data fusion to 3D scene reconstruction. Specific topics include object detection, tracking, segmentation, multi-view fusion, dynamic environment perception, real-time high-precision map updates, and 3D scene understanding and reconstruction. Additionally, the session will address challenges related to system robustness, real-time performance, and adaptability in complex driving environments, exploring the feasibility and prospects of various solution strategies. The aim of this session is to provide a platform for experts and practitioners from academia and industry to foster technical exchange and collaborative innovation, collectively advancing autonomous driving technology.

Scope of Submissions includes but is not limited to:

  • Multi-sensor Fusion

  • Object Detection and Tracking

  • Scene Understanding and Segmentation

  • Collaborative Perception

  • Transfer Learning

  • Reinforcement Learning

  • End-to-End Learning

  • High-Precision Map Construction

  • 3D Scene Reconstruction

  • Simulation Testing and Validation

Submission Deadline: March 31, 2025


Special Session Ⅴ: Deep Anomaly Detection Methods and Their Applications

Session Chair: Prof. Hongjie Xing, Hebei University, China

Keywords: Deep Anomaly Detection; Time Series Anomaly Detection; Weakly Supervised Video Anomaly Detection; Deep One-class Classification; Deep Learning

Information: The goal of anomaly detection is to identify anomalous patterns within normal patterns. Recently, deep anomaly detection methods have achieved excellent performance in handling large-scale complex data. However, with the rapid increase in data volume, there is an increasing demand for deep anomaly detection models, constantly requiring new technologies and improvements to existing ones.

The topics of this special session will include but are not limited to:

  • Deep anomaly detection

  • Time series anomaly detection

  • Weakly supervised video anomaly detection

  • Graph neural network based anomaly detection

  • Network intrusion detection

  • Machinery fault diagnosis

  • Any other application areas that employs deep anomaly detection

Submission Deadline: January 20, 2025


Special Session Ⅵ: Advances in Human Motion and Activity Analysis

Session Chairs: Prof. Xianglei Xing, Harbin Engineering University, China

Assoc. Prof. Mingliang Gao, Shandong University of Technology, China

Keywords: Gait Recognition; People Counting; Pedestrian Tracking; Behavior Analysis; Trajectory Prediction; Generative Models

Information: This session explores the latest advancements in human motion and activity analysis, covering a wide range of topics such as gait recognition, people counting, pedestrian tracking, behavior analysis, trajectory prediction, and generative models. These technologies are crucial in various applications, including surveillance, smart cities, human-computer interaction, and robotics. The session will delve into innovative methods for recognizing and tracking human movement, analyzing behaviors, and predicting trajectories, as well as generating realistic human models. Researchers and practitioners from these domains will share their latest findings, techniques, and applications in a collaborative environment.

The topics of this special session will include but are not limited to:

  • Gait Recognition and Analysis: Advances in techniques for identifying and analyzing human gait patterns for security, healthcare, and other applications.

  • People Counting and Detection: Methods and technologies for counting individuals in crowded or dynamic environments using computer vision and machine learning.

  • Pedestrian Tracking: Techniques for tracking the movement of pedestrians in real-time, including multi-object tracking and occlusion handling.

  • Behavior Analysis and Recognition: Approaches to analyze and recognize human behavior, including emotion detection, activity classification, and action recognition.

  • Trajectory Prediction: Predicting the future movement paths of individuals or groups based on historical data and real-time observations.

  • Generative Models for Human Motion: Using generative models, such as GANs, Energy-based models and Diffusion model, to create realistic human motion for

    virtual environments, animation, and robotics.

Submission Deadline: February 24, 2025