CryO and Flagellar Motors Dataset
Cryo-electronic imaging corpus in tomography for the automatic localization of flagellar motors in bacteria.
167,000 files (tomograms and 3D annotations) — various cryo-ET microscopy formats
CC0: Public Domain
Description
This dataset contains data from cryo-electron microscopy in tomography (cryo-ET), with more than 1200 tomograms and more than 1600 manual annotations of bacterial flagellar motors. It was designed as part of a scientific 3D localization challenge at BYU University. It is a very detailed data set for training or testing models for detecting biological structures at the nanoscale.
What is this dataset for?
- Training 3D detection models on cryo-ET data in molecular biology
- Automatically locate subcellular structures in bacterial cells
- Analyzing the morphology of flagellar motors in the context of biomedical research
Can it be enriched or improved?
Yes. The dataset can be enriched by methods of automatic 3D segmentation, classification of structures or complementary labeling (type of bacteria, orientation, etc.). It can also be used as a basis for super-resolution approaches or AI-assisted 3D reconstruction approaches.
🔎 In summary
🧠 Recommended for
- Biophysics researchers
- Microscopy experts
- Biomedical AI engineers
🔧 Compatible tools
- PyTorch3D
- MONAI
- Amira
- ChimeraX
- Python + Numpy/Scikit-Image
💡 Tip
For the best results, use 3D data augmentation methods before training your detection or segmentation models.
Frequently Asked Questions
What exactly does an annotation represent in this dataset?
Each annotation indicates the location of a flagellar motor identified manually in a 3D cryo-ET tomogram.
Can this data be used for 3D reconstruction?
Yes, tomograms make it possible to reconstruct bacterial 3D volumes and to locate the structures of interest there.
Is this dataset suitable for a classification task?
Indirectly. It is primarily designed for localization, but annotated regions can be used as a basis for classification.
