CEM500K - A large-scale heterogeneous unlabeled cellular electron microscopy image dataset for deep learning.

REL 2020-12-11 2020-12-21 2022-01-26 CEM500K - A large-scale heterogeneous unlabeled cellular electron microscopy image dataset for deep learning. 0000-0001-7982-6494 Kedar Narayan FNL, NCI, NIH Frederick MD United States 21701 0000-0001-7982-6494 Kedar Narayan FNL, NCI, NIH Frederick MD United States 21701 Conrad RW Narayan K 16.6 10.6019/EMPIAR-10592 FIB-SEM Conrad RW Narayan K CEM500K - A large-scale heterogeneous unlabeled cellular electron microscopy image dataset for deep learning. bioRxiv bioRxiv United States 2020 English 10.1101/2020.12.11.421792 Conrad RW Narayan K CEM500K, a large-scale heterogeneous unlabeled cellular electron microscopy image dataset for deep learning eLife eLife United States 2050-084X 2020 English 10.7554/eLife.65894.sa2 CEM500K /data/data micrographs - single frame TIFF TIFF 496544 1 UNSIGNED BYTE 224 variable 224 variable

CEM500K is a collection of 496,544 images (224x224 unsigned 8 bit .tiff) of cellular EM images, curated specifically for deep learning applications. See for more details: https://doi.org/10.1101/2020.12.11.421792 data/metadata/cem500k_image_metadata.csv file contains the details of each image file in CEM500K: image_name, source_experiment, doi, imaging_mode, organism, tissue ResNet50 weights pretrained on CEM500K using MoCoV2 algorithm in PyTorch format are stored in data/pretrained_models/cem500k_mocov2_resnet50_200ep_pth.tar