Skip to content

Latest commit

 

History

History
9 lines (5 loc) · 1.15 KB

supported_datasets.md

File metadata and controls

9 lines (5 loc) · 1.15 KB

The list of supported datasets is as follows

  1. CMU-MOSEI: CMU-MOSEI contains a large number of in-the-wild videos annotated for sentiment and emotions. The annotations follow consensus-based online perceptions. Alongside the value of this dataset for sentiment and emotion recognition, it is also suitable for representation learning due to large size of the dataset.

  2. CMU-MOSI: CMU-MOSI is a standard benchmark for multimodal sentiment analysis. It is specially suited to train and test multimodal models, since most of the newest works in multimodal temporal data use this dataset in their papers.

  3. POM: POM is a dataset for analysis of persuasion on online social media. It has annotations for personality and sentiment as well, which makes it very compelling for large number of tasks.

For downloading raw datasets (raw video, audio and text), please refer to link: http://immortal.multicomp.cs.cmu.edu/raw_datasets/. It is highly suggested that you do not use raw data unless you are running behavioral studies or extracting your own features. If you are extracting new features, you can create your own computational sequences and use the SDK.