Revisiting Classical Bagging with Modern Transfer Learning for On-the-fly Disaster Damage Detector

Abstract

Automatic post-disaster damage detection using aerial imagery is crucial for quick assessment of damage caused by disaster and development of a recovery plan. The main problem preventing us from creating an applicable model in practice is that damaged (positive) examples we are trying to detect are much harder to obtain than undamaged (negative) examples, especially in short time. In this paper, we revisit the classical bootstrap aggregating approach in the context of modern transfer learning for data-efficient disaster damage detection. Unlike previous classical ensemble learning articles, our work points out the effectiveness of simple bagging in deep transfer learning that has been underestimated in the context of imbalanced classification. Benchmark results on the AIST Building Change Detection dataset show that our approach significantly outperforms existing methodologies, including the recently proposed disentanglement learning.

Accepted at the 2019 NeurIPS Workshop on Artificial Intelligence for Humanitarian Assistance and Disaster Response(AI+HADR 2019)

References in corpus (8)

• DeepGlobe 2018: A Challenge to Parse the Earth through Satellite Images
• Using Pre-Training Can Improve Model Robustness and Uncertainty
• Deep Anomaly Detection with Outlier Exposure
• SpaceNet: A Remote Sensing Dataset and Challenge Series
• A Closer Look at Few-shot Classification
• Multi-Level Variational Autoencoder: Learning Disentangled Representations from Grouped Observations
• LSTD: A Low-Shot Transfer Detector for Object Detection
• From Satellite Imagery to Disaster Insights

Cited by in corpus (2)

• Contrastive Multiview Coding with Electro-optics for SAR Semantic Segmentation
• Multi-class segmentation under severe class imbalance: A case study in roof damage assessment