• Stars
    star
    199
  • Rank 196,105 (Top 4 %)
  • Language
    Python
  • License
    BSD 3-Clause "New...
  • Created over 7 years ago
  • Updated over 6 years ago

Reviews

There are no reviews yet. Be the first to send feedback to the community and the maintainers!

Repository Details

Implemented and improved the iTracker model proposed in the paper "Eye Tracking for Everyone"

Eye Tracker

Implemented and improved the iTracker model proposed in the paper Eye Tracking for Everyone.

Figure 1: iTracker architecture

Figure 2: modified iTracker architecture

Figures 1 and 2 show the architectures of the iTracker model and the modified model. The only difference between the modified model and the iTracker model is that we concatenate the face layer FC-F1 and face mask layer FC-FG1 first, after applying a fully connected layer FC-F2, we then concatenate the eye layer FC-E1 and FC-F2 layer. We claim that this modified architecture is superior to the iTracker architecture. Intuitively, concatenating the face mask information together with the eye information may confuse the model since the face mask information is irrelevant to the eye information. Even though the iTracker model succeeded to learn this fact from the data, the modified model outperforms the iTracker model by explictlying encoded with this knowledge. In experiments, the modified model converged faster (28 epochs vs. 40+ epochs) and achieved better validation error (2.19 cm vs. 2.514 cm). The iTracker model was implemented in itracker.py and the modified one was implemented in itracker_adv.py. Note that a smaller dataset (i.e., a subset of the full dataset in the original paper) was used in experiments and no data augmentation was applied. This smaller dataset contains 48,000 training samples and 5,000 validation samples. You can download this smaller dataset here.

Get started

To train the model: run python itracker_adv.py --train -i input_data -sm saved_model

To test the trained model: run python itracker_adv.py -i input_data -lm saved_model

You can find a pretrained (on the smaller dataset) model under the pretrained_models/itracker_adv/ folder.

FAQ

  1. What are the datasets?

The original dataset comes from the GazeCapture project. The dataset involves over 1400 subjects and results in more than 2 million face images. Due to the limitation of computation power, a much smaller dataset with 48000 training samples and 5000 validation samples was used here. Each sample contains 5 items: face, left eye, right eye, face mask and labels.

Other implementations

For pytorch implementations, see GazeCapture.

More Repositories

1

IDGL

Code & data accompanying the NeurIPS 2020 paper "Iterative Deep Graph Learning for Graph Neural Networks: Better and Robust Node Embeddings".
Python
198
star
2

BAMnet

Code & data accompanying the NAACL 2019 paper "Bidirectional Attentive Memory Networks for Question Answering over Knowledge Bases"
Python
173
star
3

KATE

Code & data accompanying the KDD 2017 paper "KATE: K-Competitive Autoencoder for Text"
Python
142
star
4

RL-based-Graph2Seq-for-NQG

Code & data accompanying the ICLR 2020 paper "Reinforcement Learning Based Graph-to-Sequence Model for Natural Question Generation"
Python
119
star
5

PFoodReq

Code & data accompanying the WSDM 2021 paper "Personalized Food Recommendation as Constrained Question Answering over a Large-scale Food Knowledge Graph"
Python
57
star
6

Graph2Seq-for-KGQG

Code & data accompanying the paper "Toward Subgraph Guided Knowledge Graph Question Generation with Graph Neural Networks"
Python
42
star
7

GraphFlow

Code & data accompanying the IJCAI 2020 paper "GraphFlow: Exploiting Conversation Flow with Graph Neural Networks for Conversational Machine Comprehension"
Python
35
star
8

AI-Papers

Favorite AI papers
17
star
9

rUDP

reliable UDP
C++
9
star
10

KDDCUP2016

Python
8
star
11

CourseNow

一个用python实现的针对电子科技大学网上选课系统的选课脚本,可实现cookie登录,预存课程,多线程选课功能……(项目来源于YaodongZhao,目前只对通信学院开放,不断完善中)
Python
7
star
12

WirelessShow

一个用python实现的图形绘制软件,可接收串口传递的加速度数据并通过计算绘制出图形轨迹。
Python
5
star
13

RenrenDataRepo

a renren.com website based user datas capture and analysis project using python 2.7
Python
4
star
14

winHook

a windows-based hook script for logging your mouse and keyboard events
Python
3
star
15

AcademicWeb

Hugo's academic homepage.
HTML
2
star
16

denoiser

Code accompanying the paper “Fast Voxel-based Surface Propagation Method for Outlier Removal”
C++
2
star
17

DVRouting

stimulating distance vector routing algorithm
C++
2
star
18

BlogWeb

Hugo's personal blog site.
HTML
1
star
19

OS

A simple operating system written in python.
Python
1
star
20

Recommender

realizing algorithms of recommenders. test & exploration...
Python
1
star
21

peoplefinder

Python
1
star
22

algorithms-in-python

Algorithm study notes in python
Python
1
star
23

TTE

C++
1
star
24

Programming-Language

Java
1
star
25

Hugo-CV

Hugo's CV
TeX
1
star
26

wirelessShow_lowerComputer

a lower computer program based on STM32f4 MPU cooperating with WirelessShow repo which is an upper computer program
C
1
star
27

ML-DM-Study-Notes

ML and DM study notes
Python
1
star
28

WebGL-Study-Notes

WebGL study notes
JavaScript
1
star
29

Linux-Programming

Linux programming example codes
C
1
star
30

DataAnalysis

A general framework for analysis of user-item data sets, which may provide suggestions for recommendation systems.
Python
1
star
31

Freeland

Hugo's personal homepage.
CSS
1
star
32

authentication

a realization of OICQ based on MFC directly communicating between server and client.
C++
1
star