Rating Image Aesthetics Using Deep Learning

Lu, Xin; Lin, Zhe; Jin, Hailin; Yang, Jianchao; Wang, James

doi:10.1109/tmm.2015.2477040

Cited by 320 publications

(447 citation statements)

References 21 publications

Supporting

Mentioning

439

Contrasting

Unclassified

Order By: Relevance

“…CNN is powerful in solving most computer vision based tasks [18][19][20][21][22] such as object recognition [23] and classification [24]. Classifying at faster rate on a huge dataset is a complicated problem.…”

Section: Literature Reviewmentioning

confidence: 99%

Indian Classical Dance Action Identification and Classification with Convolutional Neural Networks

Kishore

Kumar

et al. 2018

Advances in Multimedia

View full text Add to dashboard Cite

Extracting and recognizing complex human movements from unconstrained online/offline video sequence is a challenging task in computer vision. This paper proposes the classification of Indian classical dance actions using a powerful artificial intelligence tool: convolutional neural networks (CNN). In this work, human action recognition on Indian classical dance videos is performed on recordings from both offline (controlled recording) and online (live performances, YouTube) data. The offline data is created with ten different subjects performing 200 familiar dance mudras/poses from different Indian classical dance forms under various background environments. The online dance data is collected from YouTube for ten different subjects. Each dance pose is occupied for 60 frames or images in a video in both the cases. CNN training is performed with 8 different sample sizes, each consisting of multiple sets of subjects. The remaining 2 samples are used for testing the trained CNN. Different CNN architectures were designed and tested with our data to obtain a better accuracy in recognition. We achieved a 93.33% recognition rate compared to other classifier models reported on the same dataset.

show abstract

Section: Literature Reviewmentioning

confidence: 99%

Indian Classical Dance Action Identification and Classification with Convolutional Neural Networks

Kishore

Kumar

et al. 2018

Advances in Multimedia

View full text Add to dashboard Cite

show abstract

“…Many computer vision papers have tried to quantify and measure the aesthetic quality of images [13,14,2]. Yet this aspect of an image is subjectively derived and aesthetic values of an image will vary from subject to subject.…”

Section: Computing Image Aesthetics and Uniqueness Of Image Semanticsmentioning

confidence: 99%

Image Aesthetics and Content in Selecting Memorable Keyframes from Lifelogs

Smeaton

2018

MultiMedia Modeling

View full text Add to dashboard Cite

Abstract. Visual lifelogging using wearable cameras accumulates large amounts of image data. To make them useful they are typically structured into events corresponding to episodes which occur during the wearer's day. These events can be represented as a visual storyboard, a collection of chronologically ordered images which summarise the day's happenings. In previous work, little attention has been paid to how to select the representative keyframes for a lifelogged event, apart from the fact that the image should be of good quality in terms of absence of blurring, motion artifacts, etc. In this paper we look at image aesthetics as a characteristic of wearable camera images. We show how this can be used in combination with content analysis and temporal offsets, to offer new ways for automatically selecting wearable camera keyframes. In this paper we implement several variations of the keyframe selection method and illustrate how it works using a publicly-available lifelog dataset.

show abstract

“…), or more computation-based approaches such as image descriptors and convolutional neural networks linking visual features to expected classifications (Datta, Joshi, Li, & Wang, 2006;Lu, Lin, Jin, Yang, & Wang, 2014;Marchesotti, Perronnin, Larlus, & Csurka, 2011;Romero, Machado, Carballal, & Santos, 2012). The aesthetic classifier used in this paper extracts measures of orientation distribution, curvature distribution, HSB color distribution (Hue, Saturation, Brightness), and reflectional symmetry on cardinal and diagonal axes.…”

Section: Training Of the Aesthetic Classification Systemmentioning

confidence: 99%

From Computational Aesthetic Prediction for Images to Films and Online Videos

Lemarchand

2017

AVANT

View full text Add to dashboard Cite

In the last decade, creating and sharing videos online has become a mainstream movement and has led to some creators generating one personal video per day, also called daily vlogging. Although robust solutions exist to suggest photographs based on aesthetic criteria, the rising number of online videos created and watched means that such recommendation systems are required more than ever for videos. The main purpose of this paper is to transfer the skill of computational aesthetic classification of photographs to videos while developing new ways of investigating video creation. Using a dataset of photographs rated on aesthetic criteria by an internet community and recently developed feature extraction algorithms, the computational aesthetic classifier is capable of state-of-the-art photograph classification depending on aesthetic preferences learnt from people's ratings. On a test set of YouTube videos, the same system then displays satisfying aesthetic classification results that consist of an attempt to match the provided human aesthetic quality ratings. Achieving a transfer of skill from photograph to video classification, the computational classifier is used to analyze the evolution of aesthetics in feature films; this highlighted the aesthetic classifier's visual preferences and caused some interesting patterns to emerge that were related to filmmakers' decisions. Aesthetic classification makes it possible to observe the evolution of aesthetics over the careers of daily content creators thanks to their abundant and regular online video content. It can aid the investigation into the impact of aesthetics on the popularity of online videos using the available metadata about the internet audience's appreciation. This can also provide a new tool for video content creators to assess their work and assist them in the production of content of higher aesthetic quality.

show abstract

Rating Image Aesthetics Using Deep Learning

Cited by 320 publications

References 21 publications

Indian Classical Dance Action Identification and Classification with Convolutional Neural Networks

Indian Classical Dance Action Identification and Classification with Convolutional Neural Networks

Image Aesthetics and Content in Selecting Memorable Keyframes from Lifelogs

From Computational Aesthetic Prediction for Images to Films and Online Videos

Contact Info

Product

Resources

About