Deep Semi-Supervised Multi-Label image Classification

Abstract

The multi-label image classification (MLC) is the process that aims to firstly learn from training set of images, where each one can belong to multiple classes and so after be able to predict more than one class label simultaneously for a new tested image. This process suffers from several problems such as overlapping meaning that may contain image labels. In this thesis, we present a Deep Semi-supervised multi-label image classification method that compose modules: CAE module and ResNet module . The first module consist of a CAE model that is used to extract the features of images. The second one, is a ResNet is used to classify the extracted features.The proposed model has been trained on public benchmark dataset and it achieves better results compared to state of the art.

The multi-label image classification (MLC) is the process that aims to firstly learn from training set of images, where each one can belong to multiple classes and so after be able to predict more than one class label simultaneously for a new tested image. This process suffers from several problems such as overlapping meaning that may contain image labels. In this thesis, we present a Deep Semi-supervised multi-label image classification method that compose modules: CAE module and ResNet module . The first module consist of a CAE model that is used to extract the features of images. The second one, is a ResNet is used to classify the extracted features.The proposed model has been trained on public benchmark dataset and it achieves better results compared to state of the art.

The multi-label image classification (MLC) is the process that aims to firstly learn from training set of images, where each one can belong to multiple classes and so after be able to predict more than one class label simultaneously for a new tested image. This process suffers from several problems such as overlapping meaning that may contain image labels. In this thesis, we present a Deep Semi-supervised multi-label image classification method that compose modules: CAE module and ResNet module . The first module consist of a CAE model that is used to extract the features of images. The second one, is a ResNet is used to classify the extracted features.The proposed model has been trained on public benchmark dataset and it achieves better results compared to state of the art.