Highly Accurate Gaze Estimation Using a Consumer RGB-D Sensor

Authors: Reza Shoja Ghiass, Ognjen Arandjelovic

IJCAI 2016 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Using EYEDIAP, the standard public dataset for the evaluation of gaze estimation algorithms from RGB-D data, we demonstrate that our method greatly outperforms the state of the art.
Researcher Affiliation Academia Reza Shoja Ghiass and Ognjen Arandjelovi c Universit e Laval, Qu ebec (QC) G1V 0A6, Canada University of St Andrews, St Andrews KY16 9SX, United Kingdom
Pseudocode No The paper describes its algorithms using mathematical equations and textual explanations, but does not include any structured pseudocode or algorithm blocks.
Open Source Code No The paper does not contain any statements about making its source code publicly available or provide a link to a code repository.
Open Datasets Yes For the evaluation of the proposed method and its comparison with the state of the art we adopted the well known EYEDIAP database [Funes Mora and Odobez, 2012] 1. It is a freely and publicly available standard benchmark for the evaluation of algorithms for gaze estimation from RGB-D data. 1The database can be downloaded from http://www.idiap.ch/ dataset/eyediap.
Dataset Splits No The paper mentions using a 'training set' and refers to testing on 'subsets' of the EYEDIAP database, but does not provide specific percentages or counts for training, validation, or test splits.
Hardware Specification No The paper mentions using 'Microsoft Kinect' as a sensor, but does not specify the computational hardware (e.g., GPU, CPU models, memory) used for running the experiments or training the models.
Software Dependencies No The paper discusses the use of k-nearest neighbour (k NN) regression and adaptive linear regression (ALR) but does not provide specific software names with version numbers for any libraries or tools used.
Experiment Setup Yes To reduce the dimensionality of the representation we downsample the patches to the uniform scale of 3 5 pixels thus obtaining 15D feature vectors. We used dt = 0.01 which corresponds to the physical distance of 0.01 m. As per (15) we used the Euclidean distance... Following this work we used a sparse training set, as illustrated in Figure 2.