Evaluation of a Monocular Eye Tracking Set-Up

Overview

Evaluation of a Monocular Eye Tracking Set-Up

As part of my master thesis, I implemented a new state-of-the-art model that is based on the work of Chen et al..
For 9 calibration samples, the previous state-of-the-art performance can be improved by up to 5.44% (2.553 degrees compared to 2.7 degrees) and for 128 calibration samples, by 7% (2.418 degrees compared to 2.6 degrees). This is accomplished by (a) improving the extraction of eye features, (b) refining the fusion process of these features, (c) removing erroneous data from the MPIIFaceGaze dataset during training, and (d) optimizing the calibration method.

A software to collect own gaze data and the full gaze tracking pipeline is also available.

Results of the different models.

For the citaitions [1] - [10] please see below. "own model 1" represents the model described in the section below. "own model 2" uses the same model architecture as "own model 1" but is trained without the erroneous data, see MPIIFaceGaze section below. "own model 3" is the same as "own model 2" but with the calibrations points organized in a $\sqrt{k}\times\sqrt{k}$ grid instead of randomly on the screen.

Model

Since the feature extractors share the same weights for both eyes, it has been shown experimentally that the feature extraction process can be improved by flipping one of the eye images so that the noses of all eye images are on the same side. The main reason for this is that the images of the two eyes are more similar this way and the feature extractor can focus more on the relevant features, rather than the unimportant features, of either the left or the right eye.

The architectural improvement that has had the most impact is the improved feature fusion process of left and right eye features. Instead of simply combining the two features, they are combined using Squeeze-and-Excitation (SE) blocks. This introduces a control mechanism for the channel relationships of the extracted feature maps that the model can learn serially.

Start training by running python train.py --path_to_data=./data --validate_on_person=1 --test_on_person=0. For pretrained models, please see evaluation section.

Data

While examining and analyzing the most commonly used gaze prediction dataset, MPIIFaceGaze a subset of MPIIGaze, in detail. It was realized that some recorded data does not match the provided screen sizes. For participant 2, 7, and 10, 0.043%, 8.79%, and 0.39% of the gazes directed at the screen did not match the screen provided, respectively. The left figure below shows recorded points in the datasets that do not match the provided screen size. These false target gaze positions are also visible in the right figure below, where the gaze point that are not on the screen have a different yaw offset to the ground truth.

Results of the MPIIFaceGaze analysis

To the best of our knowledge, we are the first to address this problem of this widespread dataset, and we propose to remove all days with any errors for people 2, 7, and 10, resulting in a new dataset we call MPIIFaceGaze-. This would only reduce the dataset by about 3.2%. As shown in the first figure, see "own model 2", removing these erroneous data improves the model's overall performance.

For preprocessing MPIIFaceGaze, download the original dataset and then run python dataset/mpii_face_gaze_preprocessing.py --input_path=./MPIIFaceGaze --output_path=./data. Or download the preprocessed dataset.

To only generate the CSV files with all filenames which gaze is not on the screen, run python dataset/mpii_face_gaze_errors.py --input_path=./MPIIFaceGaze --output_path=./data. This can be run on MPIIGaze and MPIIFaceGaze, or the CSV files can be directly downloaded for MPIIGaze and MPIIFaceGaze.

Calibration

Nine calibration samples has become the norm for the comparison of different model architectures using MPIIFaceGaze. When the calibration points are organized in a $\sqrt{k}\times\sqrt{k}$ grid instead of randomly on the screen, or all in one position, the resulting person-specific calibration is more accurate. The three different ways to distribute the calibration point are compared in the figure below, also see "own model 3" in the first figure. Nine calibration samples aligned in a grid result in a lower angular error than 9 randomly positioned calibration samples.

To collect your own calibration data or dataset, please refer to gaze data collection.

Comparison of the position of the calibration samples.

Evaluation

For evaluation, the trained models are evaluated on the full MPIIFaceGaze, including the erroneous data, for a fair comparison to other approaches. Download the pretrained "own model 2" models and run python eval.py --path_to_checkpoints=./pretrained_models --path_to_data=./data to reproduce the results shown in the figure above and the table below. --grid_calibration_samples=True takes a long time to evaluate, for the ease of use the number of calibration runs is reduced to 500.

random calibration
k=9
random calibration
k=128
grid calibration
k=9
grid calibration
k=128

k=all
p00 1.780 1.676 1.760 1.674 1.668
p01 1.899 1.777 1.893 1.769 1.767
p02 1.910 1.790 1.875 1.787 1.780
p03 2.924 2.729 2.929 2.712 2.714
p04 2.355 2.239 2.346 2.229 2.229
p05 1.836 1.720 1.826 1.721 1.711
p06 2.569 2.464 2.596 2.460 2.455
p07 3.823 3.599 3.737 3.562 3.582
p08 3.778 3.508 3.637 3.501 3.484
p09 2.695 2.528 2.667 2.526 2.515
p10 3.241 3.126 3.199 3.105 3.118
p11 2.668 2.535 2.667 2.536 2.524
p12 2.204 1.877 2.131 1.882 1.848
p13 2.914 2.753 2.859 2.754 2.741
p14 2.161 2.010 2.172 2.052 1.998
mean 2.584 2.422 2.553 2.418 2.409

Bibliography

[1] Zhaokang Chen and Bertram E. Shi, “Appearance-based gaze estimation using dilated-convolutions”, Lecture Notes in Computer Science, vol. 11366, C. V. Jawahar, Hongdong Li, Greg Mori, and Konrad Schindler, Eds., pp. 309–324, 2018. DOI: 10.1007/978-3-030-20876-9_20. [Online]. Available: https://doi.org/10.1007/978-3-030-20876-9_20.
[2] ——, “Offset calibration for appearance-based gaze estimation via gaze decomposition”, in IEEE Winter Conference on Applications of Computer Vision, WACV 2020, Snowmass Village, CO, USA, March 1-5, 2020, IEEE, 2020, pp. 259–268. DOI: 10.1109/WACV45572.2020.9093419. [Online]. Available: https://doi.org/10.1109/WACV45572.2020.9093419.
[3] Tobias Fischer, Hyung Jin Chang, and Yiannis Demiris, “RT-GENE: real-time eye gaze estimation in natural environments”, in Computer Vision - ECCV 2018 - 15th European Conference, Munich, Germany, September 8-14, 2018, Proceedings, Part X, Vittorio Ferrari, Martial Hebert, Cristian Sminchisescu, and Yair Weiss, Eds., ser. Lecture Notes in Computer Science, vol. 11214, Springer, 2018, pp. 339–357. DOI: 10.1007/978-3-030-01249-6_21. [Online]. Available: https://doi.org/10.1007/978-3-030-01249-6_21.
[4] Erik Lindén, Jonas Sjöstrand, and Alexandre Proutière, “Learning to personalize in appearance-based gaze tracking”, pp. 1140–1148, 2019. DOI: 10.1109/ICCVW.2019.00145. [Online]. Available: https://doi.org/10.1109/ICCVW.2019.00145.
[5] Gang Liu, Yu Yu, Kenneth Alberto Funes Mora, and Jean-Marc Odobez, “A differential approach for gaze estimation with calibration”, in British Machine Vision Conference 2018, BMVC 2018, Newcastle, UK, September 3-6, 2018, BMVA Press, 2018, p. 235. [Online]. Available: http://bmvc2018.org/contents/papers/0792.pdf.
[6] Seonwook Park, Shalini De Mello, Pavlo Molchanov, Umar Iqbal, Otmar Hilliges, and Jan Kautz, “Few-shot adaptive gaze estimation”, pp. 9367–9376, 2019. DOI: 10.1109/ICCV.2019.00946. [Online]. Available: https://doi.org/10.1109/ICCV.2019.00946.
[7] Seonwook Park, Xucong Zhang, Andreas Bulling, and Otmar Hilliges, “Learning to find eye region landmarks for remote gaze estimation in unconstrained settings”, Bonita Sharif and Krzysztof Krejtz, Eds., 21:1–21:10, 2018. DOI: 10.1145/3204493.3204545. [Online]. Available: https://doi.org/10.1145/3204493.3204545.
[8] Yu Yu, Gang Liu, and Jean-Marc Odobez, “Improving few-shot user-specific gaze adaptation via gaze redirection synthesis”, pp. 11 937–11 946, 2019. DOI: 10.1109/CVPR.2019.01221. [Online]. Available: http://openaccess.thecvf.com/content_CVPR_2019/html/Yu_Improving_Few-Shot_User-Specific_Gaze_Adaptation_via_Gaze_Redirection_Synthesis_CVPR_2019_paper.html.
[9] Xucong Zhang, Yusuke Sugano, Mario Fritz, and Andreas Bulling, “It’s written all over your face: Full-face appearance-based gaze estimation”, pp. 2299–2308, 2017. DOI: 10.1109/CVPRW.2017.284. [Online]. Available: https://doi.org/10.1109/CVPRW.2017.284
[10] ——, “Mpiigaze: Real-world dataset and deep appearance-based gaze estimation”, IEEE Trans. Pattern Anal. Mach. Intell., vol. 41, no. 1, pp. 162–175, 2019. DOI: 10.1109/TPAMI.2017.2778103. [Online]. Available: https://doi.org/10.1109/TPAMI.2017.2778103. \

Owner
Pascal
Pascal
simple way to build the declarative and destributed data pipelines with python

unipipeline simple way to build the declarative and distributed data pipelines. Why you should use it Declarative strict config Scaffolding Fully type

aliaksandr-master 0 Jan 26, 2022
Advanced Pandas Vault — Utilities, Functions and Snippets (by @firmai).

PandasVault ⁠— Advanced Pandas Functions and Code Snippets The only Pandas utility package you would ever need. It has no exotic external dependencies

Derek Snow 374 Jan 07, 2023
apricot implements submodular optimization for the purpose of selecting subsets of massive data sets to train machine learning models quickly.

Please consider citing the manuscript if you use apricot in your academic work! You can find more thorough documentation here. apricot implements subm

Jacob Schreiber 457 Dec 20, 2022
API>local_db>AWS_RDS - Disclaimer! All data used is for educational purposes only.

APIlocal_dbAWS_RDS Disclaimer! All data used is for educational purposes only. ETL pipeline diagram. Aim of project By creating a fully working pipe

0 Apr 25, 2022
Uses MIT/MEDSL, New York Times, and US Census datasources to analyze per-county COVID-19 deaths.

Covid County Executive summary Setup Install miniconda, then in the command line, run conda create -n covid-county conda activate covid-county conda i

Ahmed Fasih 1 Dec 22, 2021
Data Science Environment Setup in single line

datascienv is package that helps your to setup your environment in single line of code with all dependency and it is also include pyforest that provide single line of import all required ml libraries

Ashish Patel 55 Dec 16, 2022
A forecasting system dedicated to smart city data

smart-city-predictions System prognostyczny dedykowany dla danych inteligentnych miast Praca inżynierska realizowana przez Michała Stawikowskiego and

Kevin Lai 1 Nov 08, 2021
Kennedy Institute of Rheumatology University of Oxford Project November 2019

TradingBot6M Kennedy Institute of Rheumatology University of Oxford Project November 2019 Run Change api.txt to binance api key: https://www.binance.c

Kannan SAR 2 Nov 16, 2021
PLStream: A Framework for Fast Polarity Labelling of Massive Data Streams

PLStream: A Framework for Fast Polarity Labelling of Massive Data Streams Motivation When dataset freshness is critical, the annotating of high speed

4 Aug 02, 2022
Using Data Science with Machine Learning techniques (ETL pipeline and ML pipeline) to classify received messages after disasters.

Using Data Science with Machine Learning techniques (ETL pipeline and ML pipeline) to classify received messages after disasters.

1 Feb 11, 2022
Import, connect and transform data into Excel

xlwings_query Import, connect and transform data into Excel. Description The concept is to apply data transformations to a main query object. When the

George Karakostas 1 Jan 19, 2022
A CLI tool to reduce the friction between data scientists by reducing git conflicts removing notebook metadata and gracefully resolving git conflicts.

databooks is a package for reducing the friction data scientists while using Jupyter notebooks, by reducing the number of git conflicts between different notebooks and assisting in the resolution of

dataroots 86 Dec 25, 2022
Provide a market analysis (R)

market-study Provide a market analysis (R) - FRENCH Produisez une étude de marché Prérequis Pour effectuer ce projet, vous devrez maîtriser la manipul

1 Feb 13, 2022
Deep universal probabilistic programming with Python and PyTorch

Getting Started | Documentation | Community | Contributing Pyro is a flexible, scalable deep probabilistic programming library built on PyTorch. Notab

7.7k Dec 30, 2022
Desafio proposto pela IGTI em seu bootcamp de Cloud Data Engineer

Desafio Modulo 4 - Cloud Data Engineer Bootcamp - IGTI Objetivos Criar infraestrutura como código Utuilizando um cluster Kubernetes na Azure Ingestão

Otacilio Filho 4 Jan 23, 2022
Exploratory Data Analysis for Employee Retention Dataset

Exploratory Data Analysis for Employee Retention Dataset Employee turn-over is a very costly problem for companies. The cost of replacing an employee

kana sudheer reddy 2 Oct 01, 2021
Demonstrate the breadth and depth of your data science skills by earning all of the Databricks Data Scientist credentials

Data Scientist Learning Plan Demonstrate the breadth and depth of your data science skills by earning all of the Databricks Data Scientist credentials

Trung-Duy Nguyen 27 Nov 01, 2022
WAL enables programmable waveform analysis.

This repro introcudes the Waveform Analysis Language (WAL). The initial paper on WAL will appear at ASPDAC'22 and can be downloaded here: https://www.

Institute for Complex Systems (ICS), Johannes Kepler University Linz 40 Dec 13, 2022
Meltano: ELT for the DataOps era. Meltano is open source, self-hosted, CLI-first, debuggable, and extensible.

Meltano is open source, self-hosted, CLI-first, debuggable, and extensible. Pipelines are code, ready to be version c

Meltano 625 Jan 02, 2023
ForecastGA is a Python tool to forecast Google Analytics data using several popular time series models.

ForecastGA is a tool that combines a couple of popular libraries, Atspy and googleanalytics, with a few enhancements.

JR Oakes 36 Jan 03, 2023