2019 IEEE/CVF International Conference on Computer Vision Workshop (ICCVW) 2019
DOI: 10.1109/iccvw.2019.00568
|View full text |Cite
|
Sign up to set email alerts
|

RITnet: Real-time Semantic Segmentation of the Eye for Gaze Tracking

Abstract: Accurate eye segmentation can improve eye-gaze estimation and support interactive computing based on visual attention; however, existing eye segmentation methods suffer from issues such as person-dependent accuracy, lack of robustness, and an inability to be run in real-time. Here, we present the RITnet model, which is a deep neural network that combines U-Net and DenseNet. RITnet is under 1 MB and achieves 95.3% accuracy on the 2019 OpenEDS Semantic Segmentation challenge. Using a GeForce GTX 1080 Ti, RITnet … Show more

Help me understand this report
View preprint versions

Search citation statements

Order By: Relevance

Paper Sections

Select...
3
1
1

Citation Types

0
38
0

Year Published

2020
2020
2024
2024

Publication Types

Select...
6
3

Relationship

0
9

Authors

Journals

citations
Cited by 66 publications
(38 citation statements)
references
References 16 publications
0
38
0
Order By: Relevance
“…Very recently some authors such as [23][24][25][26][27] have been published improvements in order to make more feasible the semantic segmentation to be used in mobile devices. Those papers also used EDS-Dataset from a Facebook competition creating a new test set reaching very competitive results.…”
Section: A Segmentation Networkmentioning
confidence: 99%
“…Very recently some authors such as [23][24][25][26][27] have been published improvements in order to make more feasible the semantic segmentation to be used in mobile devices. Those papers also used EDS-Dataset from a Facebook competition creating a new test set reaching very competitive results.…”
Section: A Segmentation Networkmentioning
confidence: 99%
“…Stated differently, the metric M is the combination of the mean intersection over union (mIOU) and model size in megabytes S. Generally, our approach achieves a competitive result, with less than half of the number of trainable parameters compared to the best result on the OpenEDS dataset as shown in Table 2. In terms of speed, our system took only 16.56 seconds while RITnet [29] took 22.75 seconds to iterate over a set of 1, 440 test images on an NVIDIA 1080Ti GPU. A comparison between our predictions and those of RITnet [29] is shown in Figure 8.…”
Section: B Evaluationmentioning
confidence: 99%
“…Perry and Fernandez [26] leveraged dilated and asymmetric convolution, while Kansal and Devanathan [27] utilized squeeze-andexcitation [16] block as well as spatial attention on channel attetion [28]. Chaudhary et al [29] presented an architecture based on DenseNet [30] and UNet [15]. They performed a lot of augmentation operations during training, such as Gaussian blur, image translation, and corruption.…”
Section: Introductionmentioning
confidence: 99%
“…Feature and model-based eye tracking systems have demonstrated to be simpler and more accurate approaches and have become the consensus solution [ 1 , 13 ]. Works applying machine learning techniques for semantic segmentation [ 14 , 15 , 16 ] or pupil center detection [ 17 , 18 ] in these controlled environments can be found. The use of convolutional neural networks (CNN) has proven to be a robust solution for pupil center detection methods in challenging images with artifacts due to poor illumination, reflections or pupil occlusion [ 17 , 18 ].…”
Section: Introductionmentioning
confidence: 99%