Automation of brain tumor segmentation in 3D magnetic resonance images (MRIs)\nis key to assess the diagnostic and treatment of the disease. In recent years,\nconvolutional neural networks (CNNs) have shown improved results in the task.\nHowever, high memory consumption is still a problem in 3D-CNNs. Moreover, most\nmethods do not include uncertainty information, which is especially critical in\nmedical diagnosis. This work studies 3D encoder-decoder architectures trained\nwith patch-based techniques to reduce memory consumption and decrease the\neffect of unbalanced data. The different trained models are then used to create\nan ensemble that leverages the properties of each model, thus increasing the\nperformance. We also introduce voxel-wise uncertainty information, both\nepistemic and aleatoric using test-time dropout (TTD) and data-augmentation\n(TTA) respectively. In addition, a hybrid approach is proposed that helps\nincrease the accuracy of the segmentation. The model and uncertainty estimation\nmeasurements proposed in this work have been used in the BraTS'20 Challenge for\ntask 1 and 3 regarding tumor segmentation and uncertainty estimation.\n