/Multimodal-Disaster-Event-Identification-from-Social-Media-Posts

We have proposed a multimodal approach. Where we first took the best unimodal for textual and visual data classification by testing and automation process. Then we fusion of the two models which can successfully classify the materials that have been damaged using the image and text data. EfficientNetB3+BERT multimodal better accuracy with 94.18%

Primary LanguageJupyter Notebook

No issues in this repository yet.