Abstract:Computational photo quality evaluation is a useful technique in many tasks of computer vision and graphics, for example, photo retaregeting, 3-D rendering, and fashion recommendation. The conventional photo quality models are designed by characterizing the pictures from all communities (e.g., “architecture” and “colorful”) indiscriminately, wherein community-specific features are not exploited explicitly. In this article, we develop a new community-aware photo quality evaluation framework. It uncovers the latent community-specific topics by a regularized latent topic model (LTM) and captures human visual quality perception by exploring multiple attributes. More specifically, given massive-scale online photographs from multiple communities, a novel ranking algorithm is proposed to measure the visual/semantic attractiveness of regions inside each photograph. Meanwhile, three attributes, namely: 1) photo quality scores; weak semantic tags; and inter-region correlations, are seamlessly and collaboratively incorporated during ranking. Subsequently, we construct the gaze shifting path (GSP) for each photograph by sequentially linking the top-ranking regions from each photograph, and an aggregation-based CNN calculates the deep representation for each GSP. Based on this, an LTM is proposed to model the GSP distribution from multiple communities in the latent space. To mitigate the overfitting problem caused by communities with very few photographs, a regularizer is incorporated into our LTM. Finally, given a test photograph, we obtain its deep GSP representation and its quality score is determined by the posterior probability of the regularized LTM. Comparative studies on four image sets have shown the competitiveness of our method. Besides, the eye-tracking experiments have demonstrated that our ranking-based GSPs are highly consistent with real human gaze movements.

Image Annotation by Latent Community Detection and Multikernel Learning.

LCMKL: latent-community and multi-kernel learning based image annotation.

A Novel Automatic Image Annotation Method Based on Co-occurrence Keywords

Labeling images by integrating sparse multiple distance learning and semantic context modeling

Community-Aware Photo Quality Evaluation by Deeply Encoding Human Perception

Automatic image annotation via local multi-label classification

The image annotation algorithm using convolutional features from intermediate layer of deep learning

Real Web Community Based Automatic Image Annotation

Effective Multi-Modal Multi-Label Learning for Automatic Image Annotation.

Two-Stage Automatic Image Annotation Based On Latent Semantic Scene Classification

A Semantic Context Model For Automatic Image Annotation

Image Annotation by Composite Kernel Learning with Group Structure

Automatic Image Annotation Exploiting Textual and Visual Saliency.

Multiple Kernel Learning Based on Weak Learner for Automatic Image Annotation.

Towards Multi-Semantic Image Annotation with Graph Regularized Exclusive Group Lasso

Image Annotation by the Multiple Kernel Learning with Group Sparsity Effect

Automatic image annotation with cooperation of concept-specific and universal visual vocabularies

Multi-label image annotation based on multi-model

Multilabel image annotation based on double-layer PLSA model.

Automatic annotation of weakly-tagged social images on flickr using latent topic discovery of multiple groups

Collaborative and Content-Based Image Labeling