no code implementations • 23 Nov 2020 • Martin Kolarik, Radim Burget, Kamil Riha
Our results show the effectiveness of Instance Normalization in the limited batch size neural network training environment.
1 code implementation • 23 Nov 2020 • Martin Kolarik, Radim Burget, Carlos M. Travieso-Gonzalez, Jan Kocica
We present a novel approach of 2D to 3D transfer learning based on mapping pre-trained 2D convolutional neural network weights into planar 3D kernels.