Wearable vision for Activity Recognition

This dataset is designed to evaluate different popular wearable computer vision systems, understanding by that cameras that are worn by a person, somehow mounted or attached to a human body, but not necessarily with the same point of view than the user.

During the acquisition of this dataset, several camera systems are worn simultaneously while performing typical indoor activities.

Sequences were acquired by 5 different users in 2 different scenarios. The goal is to acquire different perspectives of typical actions and activities performed indoors by an individual. Each trajectory performed by each user includes recordings from:

• an omnidirectional camera mounted on a helmet (helmet including IMU measurements)

• a RGB-d camera mounted on the head front

• a mobile phone camera with a wide angle lens mounted on the chest

• a wide angle camera mounted on the head front

Data available for download:

• All recordings (all users and cameras) from scenario 1 [download]

• All recordings (all users and cameras) from scenario 2 [download]

• Available annotated sequences (5 sequences) [download]