torch_brain.datasets#

Module containing base classes for creating PyTorch-compatible Datasets, as well as a library of pre-written Dataset classes for a number of brainsets.

Base Classes & Mixins#

Base classes to ease creation of PyTorch-compatible Datasets for your data.

The Dataset class is inherited by all datasets. These handle opening and accessing single datasets.
The NestedDataset class is for opening and accessing multiple datasets through a unified interface.
Mixin classes are provided to add modality-specific functionalities to the Dataset classes.

Dataset: torch_brain’s Dataset class (and its sub-classes) allow you to sample time-slices of your data. This is a major deviation from the standard torch.utils.data.Dataset, which is indexed by integers. To achieve arbitrary time-slice based access, our Dataset class is indexed by a DatasetIndex containing three attributes:

DatasetIndex(
    recording_id=...,  # The recording ID from which we want the slice
    start=...,         # Start time of the slice
    end=...,           # End time of the slice
)

Since different machine learning applications require different ways of sampling, we provide a collection of samplers which are responsible for creating these DatasetIndex objects.

See NLB Maze minimal example for an example of how to create your own Dataset subclasses.

NestedDataset: The Dataset class is designed to operate on a single dataset. However, many modern ML methods perform training over multiple datasets. For this, we provide NestedDataset that allows users to open and index through multiple datasets.

`Dataset`	PyTorch Dataset for loading time-slices of neural data recordings from HDF5 files.
`DatasetIndex`	Index for accessing a specific time interval of a recording within a `Dataset`.
`NestedDataset`	Dataset that composes multiple `Dataset` instances under a single interface.
`OpenNeuroDataset`	Base class for OpenNeuro datasets.
`SpikingDatasetMixin`	Mixin class for `torch_brain.datasets.Dataset` subclasses containing spiking data.
`CalciumImagingDatasetMixin`	Mixin class for `torch_brain.datasets.Dataset` subclasses containing calcium imaging data.
`MultiChannelDatasetMixin`	Mixin class for `torch_brain.datasets.Dataset` subclasses containing multi-channel recordings (e.g., EEG, ECoG, EMG, sEEG, etc).

Electrophysiology Datasets#

`PerichMillerPopulation2018`	Motor cortex (M1 and PMd) spiking activity and reaching kinematics from four macaques performing center-out and random target reaching tasks.
`PeiPandarinathNLB2021`	Curated spiking neural activity datasets from the Neural Latents Benchmark 2021 (NLB'21).
`ChurchlandShenoyNeural2012`	Motor cortex (M1 and PMd) spiking activity and reaching kinematics from 2 monkeys performing center-out reaching tasks with right hand.
`OdohertySabesNonhuman2017`	Motor cortex (M1 and S1) spiking activity and reaching kinematics from 2 monkeys performing random target reaching tasks with right hand.
`VollanMoserAlternating2025`	Neuropixels recordings from MEC and hippocampus in rats during spatial navigation and sleep.
`ShiraziHBNR1DS005505`	Shirazi HBN Resting State 1 (HBN-R1) iEEG Dataset (OpenNeuro DS005505).

Calcium Imaging Datasets#

AllenVisualCodingOphys2016

Two-photon calcium imaging of mouse visual cortex from the Allen Brain Observatory Visual Coding dataset, recorded during presentation of visual stimuli.

iEEG Datasets#

`Neuroprobe2025`	Neuroprobe 2025 iEEG benchmark dataset.
`KochiVisualNamingDS006914`	Kochi Visual Naming iEEG Dataset (OpenNeuro DS006914).

EEG Datasets#

KlinzingSleepDS005555

Klinzing Sleep iEEG Dataset (OpenNeuro DS005555).

PSG Datasets#

KempSleepEDF2013

Sleep-EDF Database Expanded containing 197 whole-night polysomnographic sleep recordings.