ACM MM 2026 Grand Challenge
Official Challenge Website | Rio de Janeiro, Brazil

EEG-fNIRS Handwriting-Trajectory Classification

A single-track ACM Multimedia 2026 Grand Challenge for four-class classification of imagined logographic handwriting from synchronized EEG and fNIRS signals. The benchmark emphasizes multimodal fusion, reproducible comparison, and hidden-test ranking by overall accuracy.

Register for the Challenge Dataset and Baseline
Single Track Unified task
4 Classes Logographic MI
EEG + fNIRS Full modality
ACC Ranking Hidden labels
Participation Workflow

Join in Three Steps

Participants should register first, then prepare models with the official dataset and baseline, and finally submit results during the official evaluation window.

Register

Complete the official registration form so the organizing team can contact you with challenge updates, evaluation instructions, and submission reminders.

Open registration form

Download Data and Baseline

Use the Hugging Face release to access the EEG-fNIRS handwriting-trajectory dataset and the baseline resources for model development.

Open dataset hub

Submit for Evaluation

Results submission opens on June 26, 2026 and closes on July 13, 2026. Organizers will run hidden-test scoring and rank valid submissions by overall accuracy.

Read evaluation rule
Overview

Introduction

Brain-computer interface research aims to decode neural activity into practical machine commands. Imagined handwriting trajectory decoding offers a pathway toward brain-to-text communication for users with severe paralysis.

Logographic Writing

Most trajectory decoding studies and public benchmarks focus on phonetic alphabets. This challenge centers on imagined logographic character generation.

Multimodal Sensing

EEG captures fast electrophysiological dynamics, while fNIRS captures slower but complementary hemodynamic responses.

Benchmarking

Fixed splits, synchronized EEG-fNIRS trials, baseline resources, and hidden-test ranking support reproducible comparison.

Motivation

Why This Challenge Matters

EEG-fNIRS fusion creates a realistic multimodal benchmark where methods must align heterogeneous temporal scales, noise profiles, and physiological evidence.

Fair Comparison

A standardized task definition and fixed split improve fairness across teams and make leaderboard results easier to interpret.

Hidden-Test Evaluation

Hidden labels reduce leakage and evaluation ambiguity. Organizers run the official scoring pipeline for all valid submissions.

Assistive Communication

Reliable imagined handwriting classification supports brain-to-text systems and intent-aware human-computer interaction.

Challenge Task

Single Track Classification

Classify each imagined handwriting trial into one of four logographic character classes using synchronized EEG and fNIRS recordings.

Input and Output

  • Input: synchronized EEG-fNIRS trial segments, trial IDs, and metadata from the official release.
  • Output: one predicted class label per trial in the required submission format.
  • Evaluation setting: full-modality only, using EEG and fNIRS together.

Track Rules

  • Participants should register through the official Google Form before entering the evaluation phase.
  • Self-supervised use of released unlabeled training split is allowed.
  • Any strategy requiring access to hidden test labels is prohibited.
Dataset and Baseline

Official Resources

The benchmark contains synchronized EEG and fNIRS recordings collected in an imagined handwriting motor imagery paradigm with four logographic classes.

Data Characteristics

  • Multi-session design for realistic session variability and robustness assessment.
  • EEG from multi-channel cap with standard scalp positioning.
  • fNIRS dual-wavelength measurements with hemodynamic representations.
  • Aligned event markers and standardized trial annotations.
  • Baseline resources published with the official dataset release.
Conference Policy

Presentation Policy

Please read this policy carefully before preparing a submission. This Grand Challenge follows the ACM Multimedia 2026 on-site presentation and no-show policy.

Official Policy

ACM Multimedia 2026 is an on-site event only. All papers and contributions must be presented by a physical person on-site; remote presentations will not be hosted or allowed.

Papers and contributions not presented on-site will be considered a no-show and removed from the proceedings of the conference.

More details will be provided to handle unfortunate situations in which none of the authors would be able to attend the conference physically.

Submission and Evaluation

Hidden-Test Ranking

Results submission opens on June 26, 2026 and closes on July 13, 2026. Registered teams will receive evaluation instructions from the organizing team.

Model Submission

Participants submit a runnable model package or required prediction output in the official format. The package must implement the standard inference interface that reads test trials and outputs predicted labels.

Hidden Test Execution

Organizers execute valid submissions on hidden test labels. Participants do not receive test labels and do not run official scoring locally.

Official Metric

The leaderboard is ranked solely by overall classification accuracy (ACC) on the hidden test set.

ACC = (1 / N) * sum_{i=1..N} I(y_hat_i = y_i)
  • N: number of test trials
  • y_i: ground-truth label
  • y_hat_i: predicted label
  • I(.): indicator function equal to 1 when prediction is correct, otherwise 0
Important Dates

Challenge Timeline

The dates below follow ACM Multimedia 2026 Grand Challenge schedule.

Apr 30, 2026
Data, baseline paper & code available
Jun 26, 2026
Results submission opens
Jul 13, 2026
Results submission deadline
Jul 22, 2026
Paper submission deadline
Jul 30, 2026
Paper acceptance notification
Aug 06, 2026
Camera-ready paper deadline
Nov 10–14, 2026
ACM Multimedia 2026 — Rio de Janeiro, Brazil
Ethics

Ethical Statement

The local Institutional Review Board approved the ethical conduct of this study. All procedures involving human participants follow established ethical standards, with emphasis on participant safety, privacy, and informed consent. Participants were fully informed about study purpose, procedures, and data usage policies before participation.

For questions, contact aicragmirza@gmail.com.