The GroupNormalization_Review.pdf reviews and summarizes the paper "Group Normalization" by Yuxin Wu and Kaiming He.
The Jupyter Notebook in this repository implements the layer for Group Normalization as presented in the paper "Group Normalization" by Yuxin Wu and Kaiming He
and tries to reproduce their results presented in Figure 1 with the FashionMNIST dataset.
To skip the training, replace the folders in the the root directory of the repository with the following folders containing the pre-trained weights for each model in addition.
A report describing the procedure and discussing the results is included within the Jupyter Notebook or its html version, respectively.
The contents of this repository were created in the scrope of the lecture "DeepVision - Deep Learning and Convolutional Neural Networks in Computational Vision" in winter term 2020/21 at Ulm University.