ResMem-Net: memory based deep CNN for image memorability estimation

Image memorability is a very hard problem in image processing due to its subjective nature. But due to the introduction of Deep Learning and the large availability of data and GPUs, great strides have been made in predicting the memorability of an image. In this paper, we propose a novel deep learni...

Descripción completa

Guardado en:
Detalles Bibliográficos
Autores principales: Arockia Praveen, Abdulfattah Noorwali, Duraimurugan Samiayya, Mohammad Zubair Khan, Durai Raj Vincent P M, Ali Kashif Bashir, Vinoth Alagupandi
Formato: article
Lenguaje:EN
Publicado: PeerJ Inc. 2021
Materias:
Acceso en línea:https://doaj.org/article/de7ec13f51074916a588ceccf0033b8e
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
id oai:doaj.org-article:de7ec13f51074916a588ceccf0033b8e
record_format dspace
spelling oai:doaj.org-article:de7ec13f51074916a588ceccf0033b8e2021-11-07T15:05:16ZResMem-Net: memory based deep CNN for image memorability estimation10.7717/peerj-cs.7672376-5992https://doaj.org/article/de7ec13f51074916a588ceccf0033b8e2021-11-01T00:00:00Zhttps://peerj.com/articles/cs-767.pdfhttps://peerj.com/articles/cs-767/https://doaj.org/toc/2376-5992Image memorability is a very hard problem in image processing due to its subjective nature. But due to the introduction of Deep Learning and the large availability of data and GPUs, great strides have been made in predicting the memorability of an image. In this paper, we propose a novel deep learning architecture called ResMem-Net that is a hybrid of LSTM and CNN that uses information from the hidden layers of the CNN to compute the memorability score of an image. The intermediate layers are important for predicting the output because they contain information about the intrinsic properties of the image. The proposed architecture automatically learns visual emotions and saliency, shown by the heatmaps generated using the GradRAM technique. We have also used the heatmaps and results to analyze and answer one of the most important questions in image memorability: “What makes an image memorable?”. The model is trained and evaluated using the publicly available Large-scale Image Memorability dataset (LaMem) from MIT. The results show that the model achieves a rank correlation of 0.679 and a mean squared error of 0.011, which is better than the current state-of-the-art models and is close to human consistency (p = 0.68). The proposed architecture also has a significantly low number of parameters compared to the state-of-the-art architecture, making it memory efficient and suitable for production.Arockia PraveenAbdulfattah NoorwaliDuraimurugan SamiayyaMohammad Zubair KhanDurai Raj Vincent P MAli Kashif BashirVinoth AlagupandiPeerJ Inc.articleDeep LearningImage MemorabilityVisual EmotionsSaliencyObject InterestingnessElectronic computers. Computer scienceQA75.5-76.95ENPeerJ Computer Science, Vol 7, p e767 (2021)
institution DOAJ
collection DOAJ
language EN
topic Deep Learning
Image Memorability
Visual Emotions
Saliency
Object Interestingness
Electronic computers. Computer science
QA75.5-76.95
spellingShingle Deep Learning
Image Memorability
Visual Emotions
Saliency
Object Interestingness
Electronic computers. Computer science
QA75.5-76.95
Arockia Praveen
Abdulfattah Noorwali
Duraimurugan Samiayya
Mohammad Zubair Khan
Durai Raj Vincent P M
Ali Kashif Bashir
Vinoth Alagupandi
ResMem-Net: memory based deep CNN for image memorability estimation
description Image memorability is a very hard problem in image processing due to its subjective nature. But due to the introduction of Deep Learning and the large availability of data and GPUs, great strides have been made in predicting the memorability of an image. In this paper, we propose a novel deep learning architecture called ResMem-Net that is a hybrid of LSTM and CNN that uses information from the hidden layers of the CNN to compute the memorability score of an image. The intermediate layers are important for predicting the output because they contain information about the intrinsic properties of the image. The proposed architecture automatically learns visual emotions and saliency, shown by the heatmaps generated using the GradRAM technique. We have also used the heatmaps and results to analyze and answer one of the most important questions in image memorability: “What makes an image memorable?”. The model is trained and evaluated using the publicly available Large-scale Image Memorability dataset (LaMem) from MIT. The results show that the model achieves a rank correlation of 0.679 and a mean squared error of 0.011, which is better than the current state-of-the-art models and is close to human consistency (p = 0.68). The proposed architecture also has a significantly low number of parameters compared to the state-of-the-art architecture, making it memory efficient and suitable for production.
format article
author Arockia Praveen
Abdulfattah Noorwali
Duraimurugan Samiayya
Mohammad Zubair Khan
Durai Raj Vincent P M
Ali Kashif Bashir
Vinoth Alagupandi
author_facet Arockia Praveen
Abdulfattah Noorwali
Duraimurugan Samiayya
Mohammad Zubair Khan
Durai Raj Vincent P M
Ali Kashif Bashir
Vinoth Alagupandi
author_sort Arockia Praveen
title ResMem-Net: memory based deep CNN for image memorability estimation
title_short ResMem-Net: memory based deep CNN for image memorability estimation
title_full ResMem-Net: memory based deep CNN for image memorability estimation
title_fullStr ResMem-Net: memory based deep CNN for image memorability estimation
title_full_unstemmed ResMem-Net: memory based deep CNN for image memorability estimation
title_sort resmem-net: memory based deep cnn for image memorability estimation
publisher PeerJ Inc.
publishDate 2021
url https://doaj.org/article/de7ec13f51074916a588ceccf0033b8e
work_keys_str_mv AT arockiapraveen resmemnetmemorybaseddeepcnnforimagememorabilityestimation
AT abdulfattahnoorwali resmemnetmemorybaseddeepcnnforimagememorabilityestimation
AT duraimurugansamiayya resmemnetmemorybaseddeepcnnforimagememorabilityestimation
AT mohammadzubairkhan resmemnetmemorybaseddeepcnnforimagememorabilityestimation
AT durairajvincentpm resmemnetmemorybaseddeepcnnforimagememorabilityestimation
AT alikashifbashir resmemnetmemorybaseddeepcnnforimagememorabilityestimation
AT vinothalagupandi resmemnetmemorybaseddeepcnnforimagememorabilityestimation
_version_ 1718443319301767168