CENTIVE: Motion Analysis Challenge

TelUP Human Fall Motion Dataset is an open-source research development to forecast and detect the human fall motion and analyze the motion using physical parameters and deep learning model. The dataset is collected under controlled condition in an indoor environment. Under this challenge, we provide the pose feature of single human body in the frame preprocessed with YOLOv11 Pose model. With this data, the participants are asked to put the effort to make the best model to forecast and classify the fall human motion under certain specific variables.

The TelUP Dataset is available on our Google Drive.

Download : TelUP Dataset

The pose information of TelUP Dataset is extracted using YOLOv11 to obtain 17 keypoints per frame. These keypoints are normalized and organized into sequences using a sliding window technique. The resulting sequences are then split into training and testing sets based on subject.
Finally, the data is used to train and evaluate multitask deep learning models for simultaneous forecasting and classification.

1. ⁠Design and train advanced ML/DL models capable of forecasting task and classification task. Only one model is required. One model for two tasks. The forecasting task is evaluated by MPJPE to calculate the spatial distance from ground truth to predicted key points and MPJVE is used to evaluate the smoothness of the prediction with respect to first derivative.
2. Push the limits of current DL models to make a better prediction based on the evaluation metrics.
3. Analyze the features and give a solid discussion.
4. For the purpose of this challenge, participants are provided with the official source code implementation. The code can be accessed and downloaded through the following GitHub repository:
   https://github.com/AndiDemon-Lab/HumanFallForecasting

By the end of the challenge, participants will have developed models capable of multitask environment with forecasting and classification which is the most used tasks in the DL implementation. This would not only advance the field of human-computer interaction but also improve the potential applications of fall prevention system for individuals or public used.

Why Join?

This challenge is more than a competition, it is an opportunity to:

1. ⁠Advance frontier research in human-computer interaction and artificial intelligence
2. Real-world impact, promoting the public safety with the application of falling prevention in public area
3. Collaborate with a global community with a very specific research field such as Human-AI computing, Human activity and behavior computing, etc.

All participants/teams are required to submit a paper through the CENTIVE submission system using the official CENTIVE 2025 format. Please select the Challenge track and choose Motion Analysis as the subject. All accepted papers must be presented at CENTIVE 2025. Selected papers will be published in the CENTIVE 2025 list of journals (SINTA 2 / Scopus).

Citation Requirement
As part of this challenge, it is mandatory for all participants to include a proper citation of the following paper

“TelUP Human Fall Dataset: A Motion Forecasting Study of Human Falls”
Available at: https://ejournal.ittelkom-pwt.ac.id/index.php/infotel/article/view/1420

Failure to include this citation will result in disqualification from the challenge.

Evaluation

1. The article is evaluated by MPJPE and MPJVE for the forecasting task. While Accuracy and F1 Score is used to evaluate the classification task. These score will be used to consider the winner of this challenge. However, we will also see the method used to solve and understand the problem, as well as the writings of the article.
2. Another point will be added for the article that give a comprehensive analysis.

Award

Two challengers will be awarded for this challenge with benefit as follows:

1. Prize award Rp. 2.000.000
2. Free APC publication in CENTIVE2025 journal partners
3. Certificate and Placard