The KIT Robo-kitchen data set for the evaluation of view-based activity recognition systems

Abstract

Abstract-Human action and activity recognition from videos has attracted an increasing number of researchers in recent years. However, most of the works aim at multimedia retrieval and surveillance applications, but rarely at humanoid household robots, even though the robotic perception of human activities would allow a more natural human-robot interaction (HRI). To encourage future studies in this domain, we present in this work a novel data set specifically designed for the application in HRI scenarios. This Robo-kitchen data set consists of 14 typical kitchen activities recorded in two different stereo-camera setups, and each performed by 17 subjects. To establish a baseline for future work, we extend a state-of-the-art action recognition method to be applicable on the activity classification problem and evaluate it on the Robo-kitchen data set showing promising results.

Activity detection and recognition of daily living events

“…Good separation results among the two classes have also been observed in further action datasets, such as KIT [22], HOHA [13] and UCF [21]. To further verify this, we applied the KolmogorovSmirnov test [16] on ten training videos from the URADL action dataset [19], which is often used to determine if a dataset indeed follows a Gaussian distribution.…”

“…This has led to the development of more sophisticated algorithms, but at the cost of a high computational burden, which does not allow the deployment of these methods in realistic scenarios. Recently, ADL datasets were introduced in the literature [19], [22], depicting common activities of daily life, performed by several human subjects. Their focus is on real life scenarios, but they do not address the problem of activity detection, which requires the automatic detection of activities in time.…”

RAVEL: an annotated corpus for training robots with audiovisual abilities

“…The focus of this challenge is benchmarking the different state-of-the-art action recognition methods. Last, but not least, there are three data sets concerning the daily activities on a "kitchen" scenario namely: the KIT Robo-Kitchen Activity Data Set [33], the University of Rochester Activities of Daily Living Data Set [28] and the TUM Kitchen Data Set [36].…”

"FRCEF: The new friction reduced and coupling enhanced finger for the Awiwi hand"

“…In the most recent years of humaniod robotics, research labs that are designed to suit the robot capabilities [1] are replaced by more realistic environments or even the real world [2], [3], which brings humanoid robots much closer to work with humans. This step consequently led to a paradigm change that can be seen on various robots providing enhanced robustness [4], [5], [6], [7]: Many robots nowadays are able to deal with impacts instead of trying to avoid any collision.…”

“…The Awiwi hand is able to approach all hand postures [12] and [13], as well as to keep an object firmly grasped even during an impact [14]. 1 Furthermore, it is able to withstand the impact of a 500g hammer, while in full operation, without any damage [14]. A complex system, such as the Awiwi hand, is continuously improved in order to reach always more ambitious goals.…”