Vision-based Human Fall Classification Competition


This competition is co-located at WCCI 2020, and we understand the uncertainty due to the COVID-19. Our thoughts are with those affected by the COVID-19 outbreak. We will constantly in touch with the organizers of the main conference.

The competition is open and the finalists will be invited to present their findings in a virtual session within the main conference. The awarding ceremony will also be done during the conference.

Please, feel free to register your participation and start working on it.

Due to the pandemic circumstances and after the large extended deadline, many of the participants expressed their interest to withdraw from the competition. There were no sufficient number of final submissions to deliver a winner. Unfortunately, we declared the competition void.

The competition receives 12 registered teams, and only 2 of them submitted their final results. We express our gratitude to both of them.

Final leaderboard:

  • DeepBlueAI team (92.3% of accuracy)
  • FEIT team (less than 50% of accuracy)


Due to the increasing of the world aging phenomenon, solutions that allow assisted living enhancing the quality of life and independent living of elderly people are more relevant nowadays. Particularly, human falls are frequent in this population and it is considered a major health problem. Recently, fall classification systems have been proposed in the current research of ambient assisted living (AAL). Different technologies have been considered like wearables, ambient sensors, or mobile devices. However, pervasive issues are found in those approaches. In contrast, vision-based approaches for non-contact and less pervasive solutions have been studied and implemented. Cameras in particular are cheap and easy to adopt.

To promote research and adoption of the latter technologies, this competition aims to build a vision-based classifier for detecting different types of falls and other activities of daily living in camera recordings. For competition purposes, the classifier should have a vision-based approach using machine learning and/or deep learning techniques; and the winner will be the one that best ranks, in terms of the F1-score metric, in the final results of the leaderboard.

This competition can be interesting to the growing research community of ambient assisted living, mainly on abnormal behavioral analysis. Moreover, it is also attractive to any person interested in solving computer vision and machine learning challenging problems. On the other hand, solutions to this challenge can also be applied in other domains such as robotics, biomedical engineering, or human-computer interaction.

This competition is co-located at WCCI 2020.

Organizing team

Hiram Ponce, Lourdes Martínez-Villaseñor, Ernesto Moya-Albor, Jorge Brieva, and Karina Pérez-Daniel

Facultad de Ingeniería, Universidad Panamericana, Mexico.

See our previous competition at IJCNN 2019, click here.