Accessing Public EM Datasets
Practical guide for finding, downloading, and working with public EM datasets.
Learn with real scientific data from groundbreaking research.
These pages collect datasets from landmark connectomics studies — H01, MICrONS, FlyWire, the hemibrain, and others — with links to the data portals, the papers behind them, and the site material that uses each one.
These are how-to pages, not datasets. Start here if you have chosen a volume and now need credentials, a client, or a pipeline.
Practical guide for finding, downloading, and working with public EM datasets.
A four-level ladder from browser exploration to imagery cutouts, with the auth walkthroughs, calibration checks, and failure fixes that the tool documentation leaves out.
Seven pipeline steps from fixation to circuit analysis, with the HI-MC project's numbers and a depth link into the unit that teaches each step.
Overview of the MouseConnects HI-MC dataset, technology stack, collaborators, and scientific goals.
The volumes this site teaches from. Every number below is rendered from the dataset record itself, so the card, the timeline and the dataset page cannot disagree.
A cubic millimetre of mouse visual cortex where the same neurons were both recorded in a living animal and reconstructed in EM. It is the only large volume that lets you ask what a measured cell is wired to.
The first complete connectome of an adult animal brain: every neuron, every synapse, hierarchically annotated, and proofread by a distributed community over several years.
The NIH BRAIN CONNECTS effort to reconstruct 10 mm³ of mouse hippocampal formation — an order of magnitude beyond any volume yet completed.
A cubic millimetre of human temporal cortex, imaged by serial-section EM and reconstructed automatically. It is the largest published volume of human brain at synaptic resolution, and the only one most learners will ever meet.
A densely reconstructed half of the Drosophila central brain, imaged by FIB-SEM at isotropic resolution and the dataset on which most fly circuit analysis before FlyWire was built.
Every dataset in the catalogue, oldest first. Forty years separate the first complete nervous system from the first complete adult brain.
The first complete nervous system ever mapped: 302 neurons and about 7,000 synapses, traced by hand over more than a decade. The dataset that founded the field.
The first volume to link measured physiology to EM wiring: orientation tuning recorded in a living mouse, then the same cells reconstructed.
Serial block-face EM of mouse retina combined with functional imaging, showing that starburst amacrine cells wire asymmetrically onto direction-selective ganglion cells.
A dense reconstruction of the inner plexiform layer, and the first demonstration that distributed human annotation could scale EM reconstruction.
A saturated reconstruction: every object in a small cortical volume traced, including glia and blood vessels, not just the neurons of interest.
Anatomy and function in mouse visual cortex, showing that pyramidal cells with similar tuning are preferentially connected.
The complete brain of a larval zebrafish by serial-section EM — the first vertebrate whole brain at EM resolution, and the dataset that made whole-brain structure-and-function comparison possible in a vertebrate.
A densely reconstructed half of the Drosophila central brain, imaged by FIB-SEM at isotropic resolution and the dataset on which most fly circuit analysis before FlyWire was built.
Eight complete worm brains at different developmental stages — the only dataset that shows how a connectome changes as an animal grows, and how much varies between individuals.
A cubic millimetre of human temporal cortex, imaged by serial-section EM and reconstructed automatically. It is the largest published volume of human brain at synaptic resolution, and the only one most learners will ever meet.
Motor control circuits in the adult fly ventral nerve cord, and the dataset that introduced GridTape automated section collection.
A dense connectome of the male fly ventral nerve cord, and the counterpart to FANC — the pair is the field's best evidence on how much of a connectome is sex-specific rather than species-typical.
The NIH BRAIN CONNECTS effort to reconstruct 10 mm³ of mouse hippocampal formation — an order of magnitude beyond any volume yet completed.
Synaptic development compared across mouse and primate cortex — one of the few EM datasets built for a cross-species question.
The reconstructed female fly ventral nerve cord: 14,600 cell bodies and about 45 million synapses, with leg and wing motor neurons mapped to the muscles they drive.
The first complete connectome of an adult animal brain: every neuron, every synapse, hierarchically annotated, and proofread by a distributed community over several years.
A cubic millimetre of mouse visual cortex where the same neurons were both recorded in a living animal and reconstructed in EM. It is the only large volume that lets you ask what a measured cell is wired to.
The first connectome to join a fly's brain and ventral nerve cord in one reconstruction, closing the gap that made descending and ascending neurons untraceable.
BossDB hosts many datasets beyond these. Browse bossdb.org/projects for the rest.
Connectomics datasets vary in scope, species, and methodology. Understanding these categories helps you choose the right data for your learning goals.
Most datasets are freely available through dedicated platforms. Each dataset entry above includes direct links to data portals, original publications, and popular press coverage where available.
The gap between "the data is public" and "I have a DataFrame" is where most new teams stall. This guide closes it: a four-level ladder from browser to imagery cutouts, the CAVE token walkthrough start to finish, version pinning, byte math, and a failure-signature table for when it breaks.
Curated example notebooks for downloading connectomics data from Google, the Allen Institute, Janelia, and bossDB — the notebook collection behind the getting-started ladder.