To be held at IJCAI-ECAI 2026, Bremen, Germany, on 15-17 of August 2026
We jointly hold the second workshop and challenges for 4D Micro-Expression Analysis for Mind Reading (4DMR) at IJCAI-ECAI 2026, Bremen, Germany, on 15-17 of August 2026.
We warmly welcome your contribution and participation!
May 21 : Updated workshop submission guidelines and important dates.
May 18 : The workshop submission link is now available. Updated workshop important dates.
May 14 : Updated important dates for the workshop and challenge, as well as the workshop submission guidelines and invited speakers.
May 10 : Updated workshop important dates.
April 21 : The website for the 4DMR 2026 workshop and challenge is under construction...
We hold the 2nd 4DMR Workshop & Challenge to explore the application of 4D technologies in facial expression analysis, to be held at IJCAI-ECAI 2026, Bremen, Germany, on 15-17 of August 2026.
Humans display a vast array of emotional and cognitive states. The ability to interpret these states, often referred to as mind reading, is unparalleled in the animal kingdom and is fundamental to human social interaction and communication. A key component of mind reading is facial expression, which accounts for 55% of how we understand others' feelings and attitudes, playing a vital role in conveying essential information about mental states.

Micro-expressions (ME) are a special form of facial expressions which may occur when people try to hide their true feelings for some reasons. Unlike 2D and 3D methods, 4D analysis (3D mesh + temporal changes) excels at detecting fleeting micro-expressions.4D information can be leveraged to enhance accuracy and robustness of facial expression, effectively addressing challenges such as variations in lighting, pose, and noisy environments, making it ideal for real-world applications. Despite its promise, 4D facial expression research faces challenges that limit its progress.
This workshop aims to explore the application of 4D technologies in facial expression analysis. It will feature the inaugural 4D micro-expression recognition challenge to propel the field forward and provide a platform for researchers to benchmark their methodologies. The workshop will delve into cutting-edge techniques for both macro- and micro-expression recognition, discuss the implications of these methodologies for global communication and AI systems, and highlight practical applications in domains such as security, healthcare, and customer service. Interactive sessions with leading experts will foster deeper insights into how 4D facial expression analysis can revolutionize our understanding of human emotions and cognitive states.
This workshop uses EasyChair for managing paper submissions and the peer-review process.
Humans display a vast array of emotional and cognitive states. The ability to interpret these states, often referred to as mind reading, is unparalleled in the animal kingdom and is fundamental to human social interaction and communication. A key component of mind reading is facial expression, which accounts for 55% of how we understand others' feelings and attitudes, playing a vital role in conveying essential information about mental states.
Fig. 1. 4D micro-expressions (3D mesh + temporal changes) examples.
To date, while extensive research has been conducted on facial expressions, the advent of 4D facial expression analysis marks a transformative leap in the field. By capturing the temporal evolution of expressions in three-dimensional space, 4D analysis reveals the intricate dynamics of facial muscle movements over time. Unlike 2D and 3D methods, 4D analysis (3D mesh + temporal changes) excels at detecting fleeting micro-expressions(Figure 1), which are brief, involuntary displays of hidden emotions by incorporating multiple views and temporal information for richer and more precise data. 4D information can be leveraged to enhance accuracy and robustness of facial expression, effectively addressing challenges such as variations in lighting, pose, and noisy environments, making it ideal for real-world applications. Despite its promise, 4D facial expression research faces challenges that limit its progress. The lack of diverse and realistic datasets, particularly for spontaneous micro-expressions, constrains its applicability to practical scenarios. Moreover, the computational demands of processing the complex temporal and spatial data inherent in 4D analysis pose significant technical challenges. Existing methodologies often struggle with capturing rapid and subtle micro-expressions and adapting to real-world conditions, such as occlusions, pose variations, and noisy backgrounds. Advancing the field requires the development of innovative algorithms, efficient computational techniques, and large-scale datasets to bridge these gaps, enabling applications in healthcare, security, and education.
This workshop aims to explore the application of 4D technologies in facial expression analysis. It will feature the inaugural 4D micro-expression recognition challenge to propel the field forward and provide a platform for researchers to benchmark their methodologies. The workshop will delve into cutting-edge techniques for both macro- and micro-expression recognition, discuss the implications of these methodologies for global communication and AI systems, and highlight practical applications in domains such as security, healthcare, and customer service. Interactive sessions with leading experts will foster deeper insights into how 4D facial expression analysis can revolutionize our understanding of human emotions and cognitive states.
Note: Each paper must be presented on-site by an author/co-author at the conference.
Technical University of Munich
TUM School of Medicine and Health
TUM School of Computation, Information and Technology
Talk title
Tiny Signals, Big Feelings: Affective Intelligence from Micro-Behaviour
Human behaviour reveals itself not only in explicit and intentional speech, facial expressions, and gestures, but also in subtle micro-behaviours: fleeting facial movements, gaze shifts, posture changes, vocal nuances, interaction rhythms, physiological responses, and multimodal patterns unfolding over time.
Capturing and interpreting these signals is central to affective computing and social signal processing, particularly when the goal is to infer affective, cognitive, and social states in a human-centred way.
The talk will place multimodal micro-expression analysis within the broader landscape of computational micro-behaviour understanding considering how fine-grained behavioural cues across face, body, voice, language, physiology, and context can contribute to machine perception of emotion, intention, engagement, stress, deception, rapport, and mental well-being.
Recent advances in multimodal machine learning, self-supervised representation learning, foundation and reasoning models, temporal modelling, and reliable affect recognition will be discussed as key enablers for real-world, ecologically valid settings.
The notion of “mind reading” will be addressed both as a compelling research vision and as a term requiring scientific and ethical caution. While affective AI can reveal behavioural correlates of internal states, such systems must be designed with explicit attention to uncertainty, bias, consent, privacy, transparency, and societal impact.
Peking University, CN
TBD
Micro-expressions (MEs) are subtle, rapid, and involuntary facial movements that often occur in high-stakes scenarios or when individuals attempt to gain advantages or conceal their true emotions. Due to their extremely short duration and low intensity, MEs are difficult to detect and demand high-precision facial data. This challenge leverages the power of 4D facial analysis—capturing the temporal evolution of facial expressions in 3D space—to uncover the complex dynamics of facial muscle movements over time. Unlike traditional 2D or static 3D approaches, 4D analysis (3D mesh + temporal sequence) excels at identifying fleeting, involuntary micro-expressions by incorporating both spatial depth and motion cues. This multi-view, temporal information enriches the data and significantly improves recognition accuracy and robustness.
This challenge will be organized on the Kaggle Website. On the Kaggle website, instructions will be shared, and results from participants will be submitted and ranked. The top 3 teams will be awarded certificates if the top 3 teams submit papers and are present at the workshop.
To download the dataset used in this challenge, 4DME, please carefully read and complete the license agreement. Once completed, send the signed agreement to Mengting.Wei@oulu.fi.
The timeline for the Challenge will be organized as follows:
Agricultural Information Institute of Chinese Academy of Agricultural Sciences, CN
Zhejiang University, CN
University of Oulu, FI
University of Glasgow, UK
University of Nantes
Contact Information: