Learning with privileged information for photo aesthetic assessment

Shu, Yangyang; Li, Qian; Liu, Shaowu; Xu, Guandong

doi:10.1016/j.neucom.2020.04.142

Cited by 20 publications

(12 citation statements)

References 45 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Figure 4 shows an example of the predicted images attributes with respect to the ground truth values. [10] Multi-Task CNN 0.6890* PI-DCNN (Resnet-50) [15] Multi-Task CNN 0.7051* Chen (Resnet-50) [3] Multi-Task CNN 0.7080* Pan et al(ResNet-50) [11] Multi…”

Section: Resultsmentioning

confidence: 99%

“…With the availability of more labeled data the trend has been moved from methods based on hand crafted features to deep learning methods. Recent works have both been focused on sophisticated training loss [9,15,11,3] and more powerful features [10,13,7].…”

Section: Introductionmentioning

confidence: 99%

“…In [15] authors explored the relationship between aesthetics score and aesthetics attributes introducing the PI-DCNN: a ResNet optimized over three different loss functions: regression, ranking and a privileged information loss which rely on some domain knowledge and additional information between attributes and aesthetics.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Modeling image aesthetics through aesthetics-related attributes

Leonardi¹,

Napoletano²,

Rozza³

et al. 2021

lim

View full text Add to dashboard Cite

Automatic assessment of image aesthetics is a challenging task for the computer vision community that has a wide range of applications. The most promising state-of-the-art approaches are based on deep learning methods that jointly predict aesthetics-related attributes and aesthetics score. In this article, we propose a method that learns the aesthetics score on the basis of the prediction of aesthetics-related attributes. To this end, we extract a multi-level spatially pooled (MLSP) features set from a pretrained ImageNet network and then these features are used to train a Multi Layer Perceptron (MLP) to predict image aesthetics-related attributes. A Support Vector Regression machine (SVR) is finally used to estimate the image aesthetics score starting from the aesthetics-related attributes. Experimental results on the ”Aesthetics with Attributes Database” (AADB) demonstrate the effectiveness of our approach that outperforms the state of the art of about 5.5% in terms of Spearman’s Rankorder Correlation Coefficient (SROCC).

show abstract

Section: Resultsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Modeling image aesthetics through aesthetics-related attributes

Leonardi¹,

Napoletano²,

Rozza³

et al. 2021

lim

View full text Add to dashboard Cite

show abstract

“…Apart from the model architecture, model optimization methods are also crucial to CPA. The conventional optimization approach is to minimize the mean squared error (MSE) loss [28,64,122,155] for regression tasks. To improve the MSE loss, Murray and Gordo [94] proposed the Huber loss to improve model robustness to outliers.…”

Section: Cpa Model Optimizationmentioning

confidence: 99%

“…Moreover, multi-task learning is adopted by [55,64,122] in which auxiliary training tasks are incorporated to provide extra supervision to the models, thereby enhancing CPA performance. Models reported in [55,64,122] are trained on the AADB dataset that provides scores of aestheticsrelated attributes (e.g., colour harmony, vivid colour, good lighting) [64,122]. The model jointly outputs the prediction of the aesthetics score as well as the attribute score.…”

Section: Cpa Model Optimizationmentioning

confidence: 99%

AI-Empowered Persuasive Video Generation: A Survey

Liu¹,

Han²

2021

Preprint

View full text Add to dashboard Cite

Promotional videos are rapidly becoming a popular medium for persuading people to change their behaviours in many settings (e.g., online shopping, social enterprise initiatives). Today, such videos are often produced by professionals, which is a time-, labour-and cost-intensive undertaking. In order to produce such contents to support a large applications (e.g., e-commerce), the field of artificial intelligence (AI)-empowered persuasive video generation (AIPVG) has gained traction in recent years. This field is interdisciplinary in nature, which makes it challenging for new researchers to grasp. Currently, there is no comprehensive survey of AIPVG available. In this paper, we bridge this gap by reviewing key AI techniques that can be utilized to automatically generate persuasive videos. We offer a first-of-its-kind taxonomy which divides AIPVG into three major steps: 1) visual material understanding, which extracts information from the visual materials (VMs) relevant to the target of promotion; 2) visual storyline generation, which shortlists and arranges high-quality VMs into a sequence in order to compose a storyline with persuasive power; and 3) post-production, which involves background music generation and still image animation to enhance viewing experience. We also introduce the evaluation metrics and datasets commonly adopted in the field of AIPVG. We analyze the advantages and disadvantages of the existing works belonging to the above-mentioned steps, and discuss interesting potential future research directions.

show abstract