Visual Sentiment Analysis from Disaster Images in Social Media

Image | General | Recovery

Image-sentiment dataset is a general-disaster dataset for sentiment analysis. It contains 4,003 annotated disaster related images and is introduced for multiclass multilabel classification.

  • ML task type: Multiclass multilabel classification
  • Data Source: Social Media (Twitter and Flickr, Google API)
  • Size: 4,003 Images
  • Timespan: N/A
  • Geographical Coverage: Global
  • Baseline Information
  • Evaluated on: AlexNet, VGGNet, ResNet, Inceptionv3, DenseNet, EfficientNet
  • Metrics used: Accuracy, Precision, Recall, F1-Score
  • Results as reported in original paper: Accuracy: 83.18; Precision: 83.13; Recall: 83.04; F1-Score: 82.57; ( This result is for the best performing model VGGNet(pretrained on places+ImageNet) for multilabel classification of seven classes)

Syed Zohaib Hassan, Kashif Ahmad, Steven Hicks, P˚al Halvorsen, Ala Al-Fuqaha, Nicola Conci, and Michael Riegler. Visual sentiment analysis from disaster images in social media. Sensors, 22(10):3628, 2022.