Multi Label Image Classification on VHR Satellite Images

Abstract

Multi label Image classification using Convolutional Neural Network (CNN) is yet very
difficult when it comes to performing. However, Single Label Image Classification can be
performed easily and promisingly. As there are many categories of objects in a real world image,
it becomes difficult to label them under various categories and also due to less number of multilabel training image and high complexity.This paper proposes a multi label image classification
done on Sequential CNN model taking 4-8 layers. We have trained our model with the UC
MERCED Dataset. Our model gives 88% accuracy considering Top-3 accuracy which is
significantly good for performing Multi label image classification which itself being complex