VRBagged-Net: Ensemble Based Deep Learning Model for Disaster Event Classification

A flood is an overflow of water that swamps dry land. The gravest effects flooding are the loss human life and economic losses. An early warning these events can be very effective in minimizing Social media websites such as Twitter Facebook quite efficient dissemination information pertinent to any emergency. Users on social networking sites share both textual rich content images videos. Multimedia Evaluation Benchmark (MediaEval) offers challenges form shared tasks develop evaluate new algorithms, approaches technologies for explorations exploitations multimedia decision making real time problems. Since 2015, MediaEval has been running a task predicting several aspects through tasks, many improvements have observed. In this paper, classification framework VRBagged-Net proposed implemented classification. utilizes deep learning models Visual Geometry Group (VGG) Residual Network (ResNet), along with technique Bootstrap aggregating (Bagging). Various disaster-based datasets were selected validation framework. All belong Workshop, includes Disaster Image Retrieval from Media (DIRSM), Flood Classification (FCSM) based News Topic Disambiguation (INTD). performed encouraging well all slightly different but relevant tasks. It produces Mean Average Precision at levels 98.12, 480 93.64 DIRSM. On FCSM dataset, it F1 score 90.58. Moreover, applied dataset Image-Based (INTD), exceeds previous best result by producing evaluation 93.76. slight modification also ranked first flood-related Task Workshop 2020.