All experiments were done to classify multimodal data. These experiments are the results produced by the model: https://github.com/Damorgal/Biprojection-Multimodal-Transformer
This includes:
-
Moviescope: Dataset for movie genre classification: https://arxiv.org/abs/1908.03180
-
MM-IMDb: Dataset for movie genre classification: https://arxiv.org/abs/1702.01992
-
IEMOCAP: Dataset for emotion detection: https://sail.usc.edu/iemocap/
-
CMU_MOSEI: Dataset for emotion detection and sentiment intensity: https://www.cs.cmu.edu/~pliang/papers/dap2018_mosei.pdf
A first report was done before our research. It can be find in the TechnologicalProject file.