arXiv subjects

Igor Rodriguez

Publications and source records attributed to Igor Rodriguez.

5 recordsLinked to original sources

arXiv2026

StayStill: a large-scale 3D idle animation dataset

Idle animations are essential for virtual characters, as they convey realistic behaviour during inactive states. While automatic animation generation has been widely studied, limited attention has been given to idle motion due to the absence of dedicated training datasets. We introduce StayStill, a large-scale dataset of 3D idle animations comprising diverse motion types from 50 subjects, totalling approximately 6 hours of data. We also propose a standardised evaluation protocol for both numerical and user-based metrics as a first step towards a standardised evaluation process for future systems. To facilitate future research, we publicly release StayStill along with the evaluation code and a pre-trained baseline model that generates idle animations via transition concatenation. We believe that these contributions will enable future research on idle motion generation.

cs.GR↗

arXiv2025

CapStARE: Capsule-based Sequential Architecture for Robust and Efficient Gaze Estimation

Human gaze estimation is essential for applications such as human-computer interaction, social robotics, and assistive systems. However, achieving accurate, interpretable, and real-time performance in unconstrained environments remains challenging. Existing appearance-based methods often face trade-offs between spatial robustness, computational efficiency, and effective use of contextual information. To address this, we introduce CapStARE, a capsule-based architecture that combines a frozen ConvNeXt backbone for efficient feature extraction, capsule formation with attention-based routing for structured facial reasoning, and dual GRU decoders for lightweight sequential modeling over short-horizon observation windows. This design preserves interpretable part-whole facial relationships while improving prediction stability through local contextual consistency. Experimental results demonstrate strong performance on ETH-XGaze (3.36) and MPIIFaceGaze (2.65), while also generalizing competitively on Gaze360 (9.06), all with real-time inference (<10 ms). These findings suggest that the proposed method provides a practical and robust framework for appearance-based gaze estimation in real-world interactive environments. The related code and experimental results are publicly available at: https://github.com/toukapy/capsStare

cs.CV↗

arXiv2025

Evaluating Idle Animation Believability: a User Perspective

Animating realistic avatars requires using high quality animations for every possible state the avatar can be in. This includes actions like walking or running, but also subtle movements that convey emotions and personality. Idle animations, such as standing, breathing or looking around, are crucial for realism and believability. In games and virtual applications, these are often handcrafted or recorded with actors, but this is costly. Furthermore, recording realistic idle animations can be very complex, because the actor must not know they are being recorded in order to make genuine movements. For this reasons idle animation datasets are not widely available. Nevertheless, this paper concludes that both acted and genuine idle animations are perceived as real, and that users are not able to distinguish between them. It also states that handmade and recorded idle animations are perceived differently. These two conclusions mean that recording idle animations should be easier than it is thought to be, meaning that actors can be specifically told to act the movements, significantly simplifying the recording process. These conclusions should help future efforts to record idle animation datasets. Finally, we also publish ReActIdle, a 3 dimensional idle animation dataset containing both real and acted idle motions.

cs.HC↗

arXiv2020

Quantitative analysis of robot gesticulation behavior

Social robot capabilities, such as talking gestures, are best produced using data driven approaches to avoid being repetitive and to show trustworthiness. However, there is a lack of robust quantitative methods that allow to compare such methods beyond visual evaluation. In this paper a quantitative analysis is performed that compares two Generative Adversarial Networks based gesture generation approaches. The aim is to measure characteristics such as fidelity to the original training data, but at the same time keep track of the degree of originality of the produced gestures. Principal Coordinate Analysis and procrustes statistics are performed and a new Fr\'echet Gesture Distance is proposed by adapting the Fr\'echet Inception Distance to gestures. These three techniques are taken together to asses the fidelity/originality of the generated gestures.

cs.RO↗

arXiv2019

Learning to gesticulate by observation using a deep generative approach

The goal of the system presented in this paper is to develop a natural talking gesture generation behavior for a humanoid robot, by feeding a Generative Adversarial Network (GAN) with human talking gestures recorded by a Kinect. A direct kinematic approach is used to translate from human poses to robot joint positions. The provided videos show that the robot is able to use a wide variety of gestures, offering a non-dreary, natural expression level.

cs.RO↗