Understanding Human-Object Interaction in RGB-D videos for Human Robot Interaction

Fang, Zhiwen; Yuan, Junsong; Thalmann, Nadia Magnenat

doi:10.1145/3208159.3208192

Cited by 6 publications

(2 citation statements)

References 24 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Modified A* algorithm with improvements made on rectangular symmetry reduction and jump point search for simultaneous localization and mapping of mobile robots has been used for improving the navigational character . Fang et al used RGB_D sensors with depth information for improving the human–robot interaction in recognizing and detecting objects . Zhang et al developed and presented quadratic program method and virtual plane approach for formulating the coordinated dual arm motion with the analytical solution of head motion, which improves the efficiency of the head‐arm model of robot toward human‐like behavior generation…”

Section: Introductionmentioning

confidence: 99%

Implementation of grey wolf optimization controller for multiple humanoid navigation

Muni

Parhi

Kumar

2020

Computer Animation & Virtual

View full text Add to dashboard Cite

In this paper, grey wolf optimization controller (GWOC) is considered as a multiobjective technique for multiple humanoid navigations. Upon activation of GWOC, the humanoids mimic the group hunting behavior of grey wolves and navigate toward the target in a collision-free manner in presence of both static and dynamic hurdles. The wolves in the pack will either diverge for searching prey or converge together for attacking the prey following the best search agent (Leader Alpha). GWOC has the ability to keep the humanoid free from being trapped in local minima whereas it facilitates it to head toward global minima. GWOC provides better results as compared to other intelligent techniques because of its five characteristics that include safe boundary, protection, following, hunting, and caring. Both simulation and experimental navigation in laboratory conditions for single as well as for multiple humanoid NAOs have been carried out. From the results of simulation and experimental data, it is confirmed that GWOC provides global minima for humanoid robots in complex environments with different shaped obstacles. A Petri-net controller is considered while navigating multiple humanoids, as during multiple humanoid navigations, one humanoid robot acts as a dynamic obstacle to other humanoids. K E Y W O R D SGWOC, NAO humanoid, navigation, petri-net model INTRODUCTIONLot of research works have been performed for path navigation of mobile robots, biped robots, and industrial robots by using both classical and artificial intelligent techniques. But there is still some improvement needed to avoid the trapping of robots in local minima. Today, the main challenging task for researchers is the path navigation of humanoids having multidegrees of freedom by using computational methods. Though several researchers are working on it, still it is in virtue to obtain the best intelligent method for effective and efficient navigation. For this purpose, grey wolf optimization controller (GWOC) is considered. The humanoid NAO considered for this research work is developed by ALDEBARAN robotics France, which is having 25 degrees of freedom. 1 There are sensors for detecting the obstacles, cameras to take the images of hurdles, motors, actuators, and batteries to perform the predefined tasks. The obstacle distances and shape of the obstacles are detected by inbuilt ultrasonic sensors and camera mount on the head. The programmed GWOC activates when NAO finds any obstacle in its path while heading toward target. Several researchers have worked toward the navigation of different kinds of robots using many intelligent techniques as path navigation of robots plays a Comp Anim Virtual Worlds. 2020;31:e1919. wileyonlinelibrary.com/journal/cav

show abstract

Section: Introductionmentioning

confidence: 99%

Implementation of grey wolf optimization controller for multiple humanoid navigation

Muni

Parhi

Kumar

2020

Computer Animation & Virtual

View full text Add to dashboard Cite

show abstract

“…Human-Object Interaction (HOI) Recognition is the task of identifying how people interact with the surrounding objects from the visual appearance of the scene and it is of paramount importance to understand the content of an image. It consists of producing a set of human, action, object triplets for the input image, providing a concise representation of the image semantics that can be used in higher-level tasks like Image Captioning [1] or Human-Robot Interaction [2].…”

Section: Introductionmentioning

confidence: 99%

Zero-Shot Human-Object Interaction Recognition via Affordance Graphs

Sarullo

2020

Preprint

View full text Add to dashboard Cite

We propose a new approach for Zero-Shot Human-Object Interaction Recognition in the challenging setting that involves interactions with unseen actions (as opposed to just unseen combinations of seen actions and objects). Our approach makes use of knowledge external to the image content in the form of a graph that models affordance relations between actions and objects, i.e., whether an action can be performed on the given object or not. We propose a loss function with the aim of distilling the knowledge contained in the graph into the model, while also using the graph to regularise learnt representations by imposing a local structure on the latent space. We evaluate our approach on several datasets (including the popular HICO and HICO-DET) and show that it outperforms the current state of the art.

show abstract