scispace - formally typeset
Search or ask a question
Proceedings ArticleDOI

Detecting Mood of Aesthetically Pleasing Videos Using Deep Learning

TL;DR: In this paper, an approach which is automatically assessed the aesthetics of videos, with emphasis on detecting the mood of videos is presented. But there are some videos as well as photos which are uninteresting and of low-quality.
Abstract: Human interaction and carrier of feelings among humans are accomplished mainly through five senses, such as touch, smell, taste, audio, and Visual. Considering Visual sense, images and videos are important gradients in day-to-day life. It can elevate/depress mood of a person. Video Aesthetics improves user satisfaction. In this study of mood detection, we are demonstrating an approach which is automatically assessed the aesthetics of videos, with emphasis on detecting the mood of videos. With the rising popularity of DSLR camera, Mobile dual camera, the amount of Visual data available on the internet is expanding exponentially. Some of the videos and pictures are aesthetically pleasing and beautiful to human eyes. But there are some videos as well as photos which are uninteresting and of low-quality. This paper demonstrates a simple but powerful method to classify videos into pleasing and non-pleasing categories. Further, our aesthetic quality assessment will find the mood of aesthetically pleasing video (i.e. video representing a happy video or sad or fear, etc), that reflect on a person’s mood.
Citations
More filters
01 Jan 2009
TL;DR: Schloss and Palmer as mentioned in this paper found that color preference is more closely related to the weighted affective valence estimate (WAVE) of color-associated objects than it is to the retinal cone contrasts proposed by Hurlbert and Ling.
Abstract: An Ecological Valence Theory of Human Color Preferences Karen B Schloss (kschloss@berkeleyedu) Stephen E Palmer (palmer@cogsciberkeleyedu) Department of Psychology, University of California, Berkeley, Berkeley, CA 94720-1650 USA Abstract to objects, institutions, and events associated with those colors We then describe the Berkeley Color Project and some results that support our ecological valence theory and challenge Hurlbert and Ling’s (2007) cone-contrast theory In particular, we find that average color preferences in our data are much more closely related to the weighted affective valence estimate (WAVE) of color-associated objects than it is to the retinal cone contrasts proposed by Hurlbert and Ling Although color preference is an important aspect of human behavior, little is known about why people like some colors more than others In this paper we probe this issue both theoretically and empirically First, we discuss Hurlbert and Ling’s (2007) cone-contrast theory, which posits a physiological explanation based on opponent cone outputs and gender differences We then present an ecological valence theory that color preferences reflect people’s cumulative emotional responses to environmental objects/events strongly associated with particular colors Finally, we present data that challenge Hurlbert and Ling’s model on multiple counts and support an ecological valence approach Sensory Physiological Approach Keywords: color; aesthetics; preference; ecological valence Color preference is an important aspect of human behavior It influences a wide spectrum of decisions people make on a regular basis, including the products they buy, the clothes they wear, the way they decorate their homes and offices, and how they design their personal and professional websites, to name but a few examples One reason why color preference plays such a prominent role in decision- making is that color is among the most customizable features of man-made objects Although color is, in some sense, a superficial quality that seldom influences the practical function of artifacts, there is a good reason why clothes, cars, ipods, and carpet come in such a wide variety of colors: most people prefer to surround themselves with colors they like There has been much research on which colors the average person prefers, some of which has been published in the scientific literature (eg, Eysenck, 1941; Granger, 1955; Guilford & Smith, 1959) and some of which is no doubt locked in confidential file cabinets of the corporate world Until recently, research on color preference has primarily focused on describing which colors humans prefer without shedding light on why people prefer the colors they do This level of analysis is sufficient for designers, whose goal is to produce aesthetically pleasing products For cognitive scientists, however, who are interested in why people have the color preferences they do – and indeed, why people have color preferences at all – simply describing preferences is not enough We begin by reviewing Hurlbert and Ling’s (2007) cone- based theory of color preference, which argues that it is related to the relative activation of opponent processes derived from retinal cone responses We then introduce an ecological valence theory, which proposes that color preferences reflect people’s cumulative emotional responses Hurlbert and Ling (2007) recently introduced an explanation for human color preference based on retinal cone responses They were able to account for a large portion of the variance in average color preference for their set of test colors using linear combinations of cone contrasts – specifically, L-M and S-(L+M), where L, M, and S indicate the output of cones that are most sensitive to long, medium and short wavelengths of light, respectively – calculated for the test color relative to its neutral gray background The L-M axis roughly corresponds to a red-to-cyan dimension and the S- (L+M) axis roughly corresponds to a purple-to-chartreuse dimension (although Hurlbert and Ling refer to the former as “red-green” and the latter as “blue-yellow”) They found that both male and female preferences weighted highly positively on the S-(L+M) axis, because both prefer purpler colors over chartreuser 1 colors Weights on the L-M axis differed as a function of gender, however: females weighted positively, preferring redder colors, and males weighted negatively, preferring cyaner 1 colors Hurlbert and Ling suggested that this gender difference in L-M (red/cyan) preferences is based on a hardwired biological mechanism that evolved in the context of “hunter-gather” societies In particular, they argue that females like redder colors because their visual system has specialized to be attracted to ripe berries and fruit against a background of green foliage There are both theoretical and empirical problems with their account, however First, their theory provides no explanation for their most robust finding: that both males and females robustly prefer purpler colors to chartreuser colors Second, their theory explains why females should like redder colors, but does not explain why males should like cyaner colors Even if males never picked berries and Please forgive the odd terminology here We use “chartreuser” and “cyaner” rather than “more chartreuse” and “more cyan” because of the ambiguity the latter introduces in terms of the number of colors in these categories
References
More filters
Journal ArticleDOI
TL;DR: A generic computer vision system designed for exploiting trained deep Convolutional Neural Networks as a generic feature extractor and mixing these features with more traditional hand-crafted features is presented, demonstrating the generalizability of the proposed approach.

376 citations

Journal ArticleDOI
TL;DR: In this article, an anecological valence theory was proposed to explain color preference in humans. And the empirical test provided strong support for this theory: people like colors strongly associated with objects they like (e.g., browns with feces and rottenfood).
Abstract: Color preference is an important aspect of visual experience, butlittle is known about why people in general like some colors morethan others. Previous research suggested explanations based onbiologicaladaptations[HurlbertAC,LingYL(2007)CurrBiol17:623–625] and color-emotions [Ou L-C, Luo MR, Woodcock A, Wright A(2004) Color Res Appl 29:381–389]. In this article we articulate anecological valence theory in which color preferences arise frompeople’s average affective responses to color-associated objects.An empirical test provides strong support for this theory: Peoplelike colors strongly associated with objects they like (e.g., blueswith clear skies and clean water) and dislike colors strongly associ-ated with objects they dislike (e.g., browns with feces and rottenfood).Relativetoalternativetheories,theecologicalvalencetheoryboth fits the data better (even with fewer free parameters) andprovides a more plausible, comprehensive causal explanation ofcolor preferences.

159 citations

Journal ArticleDOI
TL;DR: In this paper, a large sample of North American college students was used to verify findings surrounding sex differences in color preferences, and to extend this realm of inquiry by looking for possible differences in colour preferences associated with sexual orientation.

121 citations

Journal ArticleDOI
TL;DR: This paper proposes an automated method of video key frame extraction using dynamic Delaunay graph clustering via an iterative edge pruning strategy that improves the video summary.

102 citations

Journal ArticleDOI
TL;DR: This paper presents a multi-scene deep learning model (MSDLM) to enable automatic aesthetic feature learning and significantly outperforms existing methods on two benchmark datasets.
Abstract: Aesthetic evaluation of images has attracted a lot of research interests recently. Previous work focused on extracting handcrafted image features or generic image descriptors to build statistical model for aesthetic evaluation. However, the effectiveness of these approaches is limited by researchers' understanding on the aesthetic rules. In this paper, we present a multi-scene deep learning model (MSDLM) to enable automatic aesthetic feature learning. This deep learning model achieves better results because it improves performance on some major problems, including limited data amount and categories, scenes dependent evaluation, unbalanced dataset, noise data etc. Major innovations are as follows. (1) We design a scene convolutional layer consist of multi-group descriptors in the network elaborately so that the model has a comprehensive learning capacity for image aesthetic. (2) We design a pre-training procedure to initialize our model. Through pre-training the multi-group descriptors discriminatively, our model can extract specific aesthetic features for various scenes, and reduce the impact of noise data when building the model. Experimental results show that our approach significantly outperforms existing methods on two benchmark datasets. Present a multi-scene deep learning model for image aesthetic evaluation.Design a scene convolutional layer containing multi-group descriptors in the network.Design a pre-training procedure to initialize the model.

84 citations


"Detecting Mood of Aesthetically Ple..." refers methods in this paper

  • ...For aesthetic evaluation, there is a use of a Multi-Scene-Deep-LearningModel (MSDLM)[1]....

    [...]