Exploring Age Estimation with Facial and Tongue Biometrics:
A Challenge for Multi-modal Medical Biometrics Dataset
IEEE International Joint Conference on Biometrics (IJCB 2026)
1-4 September 2026, Rome, Italy
Embedded Files
Overview
In this competition, we invite participants to explore age estimation through the multi-modal learning of facial and tongue biometrics using a newly constructed multi-modal medical biometrics dataset. The competition features three tracks: (i) facial-based age estimation, which challenges participants to estimate age from high-resolution facial images, (ii) tongue-based age estimation, which focuses on predicting age from human tongue images that capture subtle physiological cues, (iii) multi-modal age estimation with both facial and tongue images. By leveraging this unique cross-modality dataset, the competition encourages the development of innovative multi-modal learning strategies that address modality heterogeneity, medical noise, and representation robustness. The aim is to advance multi-modal age estimation techniques with potential applications in computer-aided disease diagnostics, healthcare monitoring, and medical biometrics. The competition will be hosted on the professional open-source platform [link].
Timeline
Competition Launch: 24 March, 2026
Registration Deadline: 7 May, 2026
Start date of phase 1: 24 March, 2026
Submission Deadline of phase 1: 15 April, 2026
Start date of phase 2: 16 April, 2026
Submission Deadline of phase 2: 8 May, 2026
Results Announcement: 11 May, 2026
IJCB Paper Submission: 30 May, 2026
How to Join the Competition?
The competition is open to everyone; However, members of the organizing teams are not eligible to participate.
Important: Please register with your institutional email on Codabench. Email addresses like gmail.com or qq.com will not be accepted. Please be reminded that the Team Leader must sign the license agreement. If you are a student, your supervisor must sign.
We will provide anonymized features of tongue and facial data with age labels from the multimodal medical biometrics dataset. License Agreement (IMPORTANT: The Team Leader must sign. If you are a student, your supervisor must sign.)
Download the license aggreement here
Figure 1 Samples of multi-modal medical biometrics dataset.
About the Datasets
We provide anonymitized features of tongue and facial data with age labels from the multimodal medical biometrics dataset. Totally 3180 pair (tongue-face) of samples are selected with 15% testing (477 images).
Two versions of our dataset are provided: MBAge-tongue and MBAge-multi.
* The tongue-only version (MBAge-tongue) introduces realistic environmental variations, encouraging the development of models that can focus on tongue-based age estimation.
* The multi-modal version (MBAge-multi) provides tongue and facial images as multi-modal data for age estimation, encouraging the development of models focusing on the multi-modal data fusion and comprehensive age estimation.
Check [link] for more details.
MBAge-tongue
Model - Feature Shape (H x W @ C)
- ResNet18 - 28 x 28 @ 128
- MobileNetV2 - 28 x 28 @ 32
- EfficientNetB0 - 28 x 28 @ 40
- DenseNet121 - 28 x 28 @ 128
MBAge-multi
(To be announced when phrase 2 is open.)
Evaluation Protocol
4 metrics are adopted for comprehensive evaluation: Mean Squared Error (MSE), 1-year Prediction Accuracy(1Y-ACC), 5-year Prediction Accuracy(5YACC), and 10-year Prediction Accuracy(10YACC).
The final score are calculated as:
Reward (Certificate+IJCB paper authorship+Cash)
We are delighted to recognize the outstanding achievements of the top three teams in this competition with the following rewards:
1. Award Certificates: All members of the winning teams will receive official certificates of achievement.
2. IJCB submission Co-Authorship: Up to three members per team (from the 1st, 2nd, and 3rd-place teams) will be invited as co-authors on a competition-related paper to be submitted to the 2026 International Joint Conference on Biometrics (IJCB2026)
3. Cash reward: For the 1st, 2nd, and 3rd prize winners, we will provide $400, $200, and $100 dollars as the reward.
Winners
(Important): We will calculate the final score of the whole competition and determine the 1st, 2nd, and 3rd prize winners by comprehensively consider the score of each phrase:
S_{final} = 0.4S_1 + 0.6S_2, where S_1 and S_2 indicate the score of phrase 1 and phrase 2.
If you only attend single phrase (1 or 2) and rank in top 3, we will also contact you to issue the certification and authorship invitation of another paper.
Organizers
Shanghai University, China
Mr. Jian Hwee Ang
University of Macau, China
University of Macau, China
The University of Osaka, Japan
Dr. Xiang Li
The University of Osaka, Japan
Chinese University of Hong Kong (Shenzhen), China
Dataset Download Links
Link of the datasets (Request Access after signing License Agreement, please mention your name and email address as in signed liscense in the message when requesting access):
MBAge-multi (Will be made available at phase 2)
Sponsors
Page updated
Google Sites
Report abuse