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.

Learning Pseudo Metrics for Semantic Image Clustering

Learning Pseudo Metric for Intelligent Multimedia Data Classification and Retrieval

Learning Based Neural Similarity Metrics for Multimedia Data Mining.

Semantic Discriminative Metric Learning for Image Similarity Measurement

Distance Metric Learning from Uncertain Side Information with Application to Automated Photo Tagging

Learning similarity for semantic images classification

KPML: A Novel Probabilistic Perspective Kernel Mahalanobis Distance Metric Learning Model for Semi-supervised Clustering

Learning Non-Metric Partial Similarity Based on Maximal Margin Criterion

SPICE: Semantic Pseudo-labeling for Image Clustering

Community-Aware Photo Quality Evaluation by Deeply Encoding Human Perception

Face Recognition Based on Local Binary Pattern and Improved Pairwise-constrained Multiple Metric Learning

Layer-Wise Feature Metric of Semantic-Pixel Matching for Few-Shot Learning

Hierarchical Semantic Segmentation using Psychometric Learning

Semantic-Enhanced Image Clustering

Collaborative Similarity Metric Learning for Semantic Image Annotation and Retrieval.

Deep discriminative image feature learning for cross-modal semantics understanding

What Makes Objects Similar: A Unified Multi-Metric Learning Approach.

Metric networks for enhanced perception of non-local semantic information

Semantic image classification using statistical local spatial relations model

A Prototypical Metric Learning Approach for Open-Set Semantic Segmentation on Remote Sensing Images

Confusion-Based Metric Learning for Regularizing Zero-Shot Image Retrieval and Clustering