- Lo
- Il
- Ma
Lounes Douar, Ilia Des, Matthew Lowe and others are enlisted in the challenge
2025 Munich Tech Arena
2025 Munich Tech Arena
Media Technology
Anthropometric Data Extraction
Head and Ear Parameters Estimation from a set of Multiple-viewpoint Images
The goal of this challenge is to extract anthropometric data [1], specifically head and ear parameters, from a series of RGBD images of a human subject. As illustrated in Fig. 1, submitted models are expected to accept a fixed number of subject images captured from predefined camera positions and output precise anthropometric measurements of the human head and ears, with particular emphasis on the pinnas.
Fig.1: Block diagram of the anthropometric feature extraction process
Provided input:
A dataset consisting of subject images along with corresponding anthropometric measurements is provided.
The subject images are captured from horizontal camera positions evenly spaced at 5-degree intervals around each subject, resulting in 72 images per subject. These RGBD images are supplied in HEIC format [2].
For evaluation purposes, the images will be organized into three different input sets:
-Full Set: All 72 images captured from every direction.
-Front-Facing Set: 36 images captured from the front half of the subject.
-Minimal Set: Only 3 images, one each from the front, left, and right views.
The order of the input images, corresponding to their camera positions relative to the subject, will remain fixed throughout the evaluation.
The anthropometric data comprises 1 measurement of the head and 5 measurements for each pinna (left and right), yielding an 11-dimensional vector in total. The specific anthropometric parameters are defined based on landmarks illustrated in Fig. 1:
Thus, the overall anthropometric vector is constructed as follows:
Expected output:
Participants must provide anthropometric measurements of both the head and the pinnas.
The submitted model should be able to handle each of the three input configurations described in the preceding section.
Evaluation Method Description:
The model will be evaluated on an undisclosed set of subjects. For each subject, the submitted model will be evaluated on 3 different sets of input images, as described in the above section. The output anthropometric vector will be compared against the ground truth anthropometric vector using the Euclidean distance:
In this equation, Sout denotes the model output anthropometric vector, Sgt the ground truth anthropometric vector, and N the number of feature dimensions.
Notably, since the evaluation is based on Euclidean distance, lower values indicate higher accuracy and therefore better performance. A reference score, computed using the average anthropometric measurements across all subjects in the dataset, serves as a lower bound for performance. While this reference does not represent an optimal solution, it provides a baseline that submitted systems are expected to surpass in order to be considered valid.
Reference:
[1] https://en.wikipedia.org/wiki/Anthropometry
[2] https://en.wikipedia.org/wiki/High_Efficiency_Image_File_Format
NDA Requirement
NDA Requirement for Anthropometric Data Extraction Track Participants
Teams that choose to participate in the Anthropometric Data Extraction track are required to submit a signed Non-Disclosure Agreement (NDA) in order to receive access to the project datasets.
Please read the following requirements and follow the instructions carefully:
Key Requirements
1. The NDA is provided by Imperial College London, the data provider for the Anthropometric Data Extraction track.
2. All team members must sign the same NDA file, with each member signing in the designated signature area.
3. The signed NDA will be forwarded and processed by Imperial College London.
4. If a team in the Anthropometric Data Extraction track has already received access to the dataset, any newly added members must also submit a signed NDA. Failure to do so will result in the revocation of the team’s data access. Please make sure all team members comply with this requirement.
Step-by-Step Instructions
Step 1: Download the NDA Template through download link
Step 2: Sign the NDA
-The NDA may be printed, signed, and scanned, or signed electronically.
-Each team member must sign the NDA individually.
Step 3: Upload the Signed NDA and Provide Your Team Email
-Log in to the Agorize platform and navigate to the Submission section.
-Please enter an active team email address in the designated field. This email will be used to deliver the dataset once your NDA has been successfully verified.
-Upload the signed NDA through the Participation Form as a single PDF file, which includes the signatures of all team members in one document.
Attached below is an example of the NDA signature page format for your reference.
Step 4: Download the Dataset
-After verification by Imperial College London, a download link will be sent to your team’s email address.
-Follow the instructions in the email: click the link, log in to the data portal, select the relevant folders, and click “Download” to retrieve the dataset.
⚠️ Important Notice:
Dataset access will only be granted after the NDA has been successfully submitted and verified by the organizing committee and Imperial College London.
Dataset access will only be granted after the NDA has been successfully submitted and verified by the organizing committee and Imperial College London.
After submitting the signed NDA on the platform, it typically takes up to one week to receive the dataset link. The Imperial College London team will send dataset access emails every Wednesday to the email addresses provided. Please plan your time accordingly.
Two-Part Evaluation Structure
The competition consists of two evaluation parts:
Part 1 – Idea Submission Evaluation
All teams are required to submit a written proposal outlining their solution approach by September 15, 2025.
👉 [Click here to view the idea template]
Submissions will be reviewed by our judging panel to ensure they meet the template and content requirements.
If your submission is approved, your team will receive full marks for Part 1, which accounts for 20% of the final score.
All teams are required to submit a written proposal outlining their solution approach by September 15, 2025.
👉 [Click here to view the idea template]
Submissions will be reviewed by our judging panel to ensure they meet the template and content requirements.
If your submission is approved, your team will receive full marks for Part 1, which accounts for 20% of the final score.
Part 2 – Final Project Evaluation
All teams are required to submit a final project approach by October 15, 2025.
The final deliverable will be evaluated by Huawei experts based on technical quality, innovation, and completeness.
This part contributes 80% of the total score.
This part contributes 80% of the total score.