DeePiCt Training and Evaluation Datasets

Deposition Overview

Datasets comprising 20 annotated Schizosaccharomyces pombe cryo-electron tomograms, with 10 tomograms acquired using a Volta phase plate (VPP) for enhanced contrast and 10 using conventional defocus. The dataset includes comprehensive annotations of ribosomes, fatty acid synthases, membranes, organelles, and cytosol. Additionally, 3 tomograms of RPE-1 cells used to train networks for actin segmentation are included. These datasets enabled the development and validation of DeePiCt, a deep learning framework combining 2D and 3D convolutional neural networks for cellular structure segmentation and macromolecular complex localization across different species and acquisition conditions.