Comparing Deep Learning Models for Image Classification in Urban Flooding

Frequent heavy rains in the city of Rio de Janeiro has been causing floods, which cause structural and economic damages. Additionally, these events require the municipal government to take quick actions in flooded areas. Detection of flooded regions through street cameras becomes crucial for this purpose. To address this issue, a Convolutional Neural Network (CNN) architecture is trained and evaluated among five different models: VGG19, InceptionV3, DenseNet, MobileNetV3 and Vision Transformer (ViT). The ViT architecture demonstrates better results than the others and serves as a good starting point for developing a more robust model. However, it also indicates that there is still room for improvement. Such elements are addressed in this work.