Where's YOUR focus: Personalized Attention

Lin, Sikun; Hui, Pei

doi:10.48550/arxiv.1802.07931

Cited by 2 publications

(5 citation statements)

References 0 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In practice however, Code available at https://github.com/nabsabraham/focal-tversky-unet it faces difficulty balancing precision and recall due to small regions-of-interest (ROI) found in medical images. Research efforts to address small ROI segmentation propose more discriminative models such as attention gated networks [5], [6]. CNNs with attention gates (AGs) focus on the target region, with respect to the classification goal, and can be trained endto-end.…”

Section: Introductionmentioning

confidence: 99%

A Novel Focal Tversky Loss Function With Improved Attention U-Net for Lesion Segmentation

Abraham

Khan

2019

2019 IEEE 16th International Symposium on Biomedical Imaging (ISBI 2019)

646

396

View full text Add to dashboard Cite

We propose a generalized focal loss function based on the Tversky index to address the issue of data imbalance in medical image segmentation. Compared to the commonly used Dice loss, our loss function achieves a better trade off between precision and recall when training on small structures such as lesions. To evaluate our loss function, we improve the attention U-Net model by incorporating an image pyramid to preserve contextual features. We experiment on the BUS 2017 dataset and ISIC 2018 dataset where lesions occupy 4.84% and 21.4% of the images area and improve segmentation accuracy when compared to the standard U-Net by 25.7% and 3.6%, respectively.

show abstract

Section: Introductionmentioning

confidence: 99%

A Novel Focal Tversky Loss Function With Improved Attention U-Net for Lesion Segmentation

Abraham

Khan

2019

2019 IEEE 16th International Symposium on Biomedical Imaging (ISBI 2019)

646

396

View full text Add to dashboard Cite

show abstract

“…However, the different persons actually focus on different areas even when they gaze at the same scene, that is, individual differences exist [8][9][10]. To model the individual visual attention, the personzalization of the salinacy map has been addressed over the past few years [11][12][13]. For distinguishing between a traditional saliency map and its personalisation, we call a universal saliency map (USM) and a personalized saliency map (PSM), respectively.…”

Section: Introductionmentioning

confidence: 99%

“…Then the gaze patterns emerging in images are quite complex and individually different, and those characteristics lead to the difficulty of the PSM prediction. For extracting the gaze patterns and tendencies, several researchers have collected eye-tracking data for thousands of images [11,12,14]. Moreover, in these researches, the simultaneous prediction of PSMs for several persons has been tried by using a multi-task convolutional neural network (multi-task CNN) [19] to compensate for the lack of data [12].…”

Section: Introductionmentioning

confidence: 99%

“…Moreover, in these researches, the simultaneous prediction of PSMs for several persons has been tried by using a multi-task convolutional neural network (multi-task CNN) [19] to compensate for the lack of data [12]. In [11,14], personalized information has been introduced into the PSM prediction model of several persons as the same training data. The prediction models adopted in these researches are based on deep learning that requires a massive amount of training data for each person.…”

Section: Introductionmentioning

confidence: 99%

“…Here, we assume that enough amount of eye-tracking data is acquired from training persons, and this is not arbitrary as there is a large-scale PSM dataset. Under this assumption, we can use the PSMs of training persons for any images since their PSMs can be predicted by using the previous researches [11,12,14]. In this research, the similarity of gaze tendencies between training persons and the target person is used for the simple method of taking a weighted average of PSMs obtained from training persons to predict PSMs of the target person.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Few-Shot Personalized Saliency Prediction with Similarity of Gaze Tendency Using Object-Based Structural Information

Moroto

Maeda

Ogawa

et al. 2022

2022 IEEE International Conference on Image Processing (ICIP)

View full text Add to dashboard Cite

This paper presents a few-shot personalized saliency prediction method with similarity of gaze tendency using object-based structural information. The personalized saliency maps (PSMs) that represent the individual visual attention can be used for analyzing the heterogeneity among personalized preferences, whereas general saliency maps ignore the individual differences. However, the PSM prediction is a difficult task since the acquisition of eye tracking data, which are needed for obtaining PSMs, gives persons a heavy burden. Then, for realizing PSM prediction with a limited amount of training data, the use of the similarity of gaze tendency between persons can be one effective way. It has been reported that human gazes are related to objects and their relative relationships that are semantic and structural information, and we focus on the integration of PSMs predicted for other persons by using object-based similarities of gaze tendency. The advantage in this paper is that we newly focus on similarities of gaze tendency for visually similar objects for solving the lack of eye tracking data by considering both semantic and structural information, simultaneously. By experimenting with the open dataset, the proposed method outperforms the state-of-the-art methods.

show abstract

Where's YOUR focus: Personalized Attention

Cited by 2 publications

References 0 publications

A Novel Focal Tversky Loss Function With Improved Attention U-Net for Lesion Segmentation

A Novel Focal Tversky Loss Function With Improved Attention U-Net for Lesion Segmentation

Few-Shot Personalized Saliency Prediction with Similarity of Gaze Tendency Using Object-Based Structural Information

Contact Info

Product

Resources

About