For many years, convolutional neural networks dominated the field of computer vision, not least in the medical field, where problems such as image segmentation were addressed by such networks as the U-Net. The arrival of self-attention-based networks to the field of computer vision through ViTs seems to have changed the trend of using standard convolutions. Throughout this work, we apply different architectures such as U-Net, ViTs and ConvMixer, to compare their performance on a medical semantic segmentation problem. All the models have been trained from scratch on the DRIVE dataset and evaluated on their private counterparts to assess which of the models performed better in the segmentation problem. Our major contribution is showing that the best-performing model (ConvMixer) is the one that shares the approach from the ViT (processing images as patches) while maintaining the foundational blocks (convolutions) from the U-Net. This mixture does not only produce better results ($DICE=0.83$) than both ViTs ($0.80$/0.077 for UNETR/SWIN-Unet) and the U-Net ($0.82$) on their own but reduces considerably the number of parameters (2.97M against 104M/27M and 31M, respectively), showing that there is no need to systematically use large models for solving image problems where smaller architectures with the optimal pieces can get better results.

The global pandemic of COVID-19 has changed our daily habits and has undoubtedly affected our smartphone usage time. This paper attempts to characterize the changes in the time of use of smartphones and their applications between the pre-lockdown and post-lockdown periods in Spain, during the first COVID-19 confinement in 2020. This study analyzes data from 1940 participants, which was obtained both from a survey and from a tracking application installed on their smartphones. We propose manifold learning techniques such as clustering, to assess, both in a quantitative and in a qualitative way, the behavioral and social effects and implications of confinement in the Spanish population. We also determine the Big Five personality traits along with addiction, Social Digital Pressure and depression indicators for every group determined by the clustering.