Visual Sentiment Analysis
Sentiment Analysis of Urban Images by Using Semantic and Deep Features
Project Goal
The present study focus on the understanding of the sentiment of urban images shared by users on social networks.
PerceptSent
Cesar R Lopes, Rodrigo Minetto, Myriam R Delgado, and Thiago H Silva. 2022. PerceptSent - Exploring Subjectivity in a Novel Dataset for Visual Sentiment Analysis. IEEE Transactions on Affective Computing - link to the paper
Data
PerceptSent Dataset: Composed of 500 images from Flickr, Instagram, and NYC311. A pool of five evaluators labelled all images as positive, slightly positive, neutral, slightly negative, and negative. Besides the sentiment opinion for each image, the dataset contains metadata concerning~ each evaluator --- age, gender, socio-economic strata, education, and psychological hints --- their perceptions, such as the presence of nature, violence, lack of maintenance, among others --- as well as independent scene objects annotated in the images.
Code
The ConvNet architectures explored in this paer, trained models are available here.
OutdoorSent
Wyverson Bonasoli de Oliveira, Leyza Baldo Dorini, Rodrigo Minetto, and Thiago H. Silva. 2020. OutdoorSent: Sentiment Analysis of Urban Outdoor Images by Using Semantic and Deep Features ACM Trans. Inf. Syst. 38, 3, Article 23 (May 2020), 28 pages. Link to the paper
Data
OutdooSent Dataset: Composed of 1,950 outdoor images from Flickr. Each of them was labeled based on the evaluation of five different volunteers. Then, each image is labeled as Positive, Neutral, or Negative.
Team
Researchers
Students
André L. Zanelatto
Cesar Rafael Lopes
Lucas Nogueira
Thiago Mildemberger
Wyverson Bonasoli de Oliveira
Acknowledgements
Grant CNPq-URBCOMP (#403260/2016-7).
Grant Fapesp-GoodWeb (#2023/00148-0).
CNPq (grants 314699/2020-1 and 310998/2020-4).
NVIDIA Corporation due to the donation of the Titan Xp GPU used for this research.
Research agencies: CAPES, CNPq, and FAPESP.