Learning Visual Sentiment Distributions via Augmented Conditional Probability Neural Network

Flickr LDL and Twitter LDL are two datasets collected based on label distribution learning for visual sentiment distributions. Flickr LDL contains 11,150 images and Twitter LDL contains 10,045 images, which are labeled by 11 and 8 viewers respectively. Each viewer can label the image with one emotion from Amusement, Awe, Contentment, Excitement, Anger, Disgust, Fear and Sadness.In both datasets, the votes from the workers are integrated to generate the ground truth label distribution for each image.

 

Download

The Readme file explains everything.

 

Datasets Presentation

 

 

Results Presentation

 

Publications

J.Yang, M Sun, X. Sun

Learning Visual Sentiment Distributions via Augmented Conditional Probability Neural Network.

AAAI Conference on Artificial Intelligence (AAAI), 2017

Document: PDF