This was my Bachelor's thesis. Fake news detection using deep learning architectures will be the focus of this graduation thesis. The dataset has been provided as part of the 2020 ICWSM data challenge task. The discussion and comparative analysis of several DL architectures for identifying fake news based on linguistic analysis of the content of news articles in the provided dataset will be followed by an approach that is augmented with graph-based analysis using graph convolutional network.

Driven by the recent event on our planet, we have concluded that the fires are a major threat to the Earth. So, our objective was finding the most efficient way for identifying fires in images. We have focused on three different approaches, trying different libraries i.e. Keras and Fastai and trying different Convolutional Networks i.e. ResNet34, VGG-16, VGG19 and self-made CNN. Also, we have experimented with pretrained models as well as models that are being trained from scratch. Moreover, we have done augmentation in order to increase the dataset and decrease the sensitivity. In the next sections, we give detailed explanation of the methods and networks, and interpret the results we got.