Spiking Heidelberg Digits and Spiking Speech Commands

In collaboration with the Electronic Vision(s) Group at the University of Heidelberg, we developed two new spiking datasets for the evaluation of spiking neural networks. The Spiking Heidelberg Digits (SHD) dataset and the Spiking Speech Command (SSC) dataset are both audio-based classification datasets for which input spikes and output labels are provided. The datasets are released under the Creative Commons Attribution 4.0 International License.


Datasets: https://zenkelab.org/datasets/
Mirrors: https://ieee-dataport.org/open-access/heidelberg-spiking-datasets and https://compneuro.net/datasets/

Conversion code: https://github.com/electronicvisions/lauscher

Leader board SHD

PublicationAccuracy %NetworkURL
Hammouamri et al. 202395.1 ± 0.3Fully connected SNN with learned delayshttps://arxiv.org/abs/2306.17670
Bittar and Garner 202294.6RSNN with adaptationhttps://www.frontiersin.org/articles/10.3389/fnins.2022.865897
Sun et al. 202392.45Feed-forward SNN with adaptive axonal delayshttps://ieeexplore.ieee.org/abstract/document/10094768
Yu et al. 202292.4Feed-forward SNN with spatio-temporal filters and attentionhttps://www.frontiersin.org/articles/10.3389/fnins.2022.1079357
Yao et al. 202191.1RSNN with temporal attentionhttps://arxiv.org/abs/2107.11711
D’Agostino et al. 202390.1% / 87.6 %Feed-forward SNN with random dendritic delays (simulation/hardware, RRAM)https://arxiv.org/abs/2312.08960
Nowotny et al. 202284.8 ± 1.5RSNNhttps://arxiv.org/abs/2212.01232
Yin et al. 202084.4RSNN with adaptionhttps://arxiv.org/abs/2005.11633
Rossbroich et al. 202283.5 ± 1.5Recurrent convolutional SNN with fluctuation-driven inithttps://iopscience.iop.org/article/10.1088/2634-4386/ac97bb
Cramer et al. 202083.2 ± 1.3RSNN with data augmentation + noise injectionhttps://doi.org/10.1109/TNNLS.2020.3044364
Perez-Nieves et al. 202182.7 ± 0.8RSNN with heterogeneous time constantshttps://www.biorxiv.org/content/10.1101/2020.12.18.423468v2
Cramer et al. 202071.4 ± 1.9RSNNhttps://doi.org/10.1109/TNNLS.2020.3044364
Cramer et al. 202048.1 ± 1.6feed-forward SNN (single hidden layer)https://doi.org/10.1109/TNNLS.2020.3044364

Please let us know if your work should be on this list.

Also please take note of the leader boards at Papers with Code for SHD https://paperswithcode.com/sota/audio-classification-on-shd and SSC https://paperswithcode.com/sota/audio-classification-on-ssc.


When using these data or the code for your work, please cite:

Cramer, B., Stradmann, Y., Schemmel, J., and Zenke, F. (2022).
The Heidelberg Spiking Data Sets for the Systematic Evaluation of Spiking Neural Networks.
IEEE Transactions on Neural Networks and Learning Systems 33, 2744–2757.


We provide two distinct classification datasets for spiking neural networks.

NameClassesSamples (train/valid/test)Parent datasetURL
SHD208156/-/2264Heidelberg Digits (HD)https://zenkelab.org/datasets/hd_audio.tar.gz
SSC3575466/9981/20382Speech Commands v0.2https://arxiv.org/abs/1804.03209

Both datasets are based on respective audio datasets. Spikes in 700 input channels were generated using Lauscher, an artificial cochlea model. The SHD consists of approximately 10000 high-quality aligned studio recordings of spoken digits from 0 to 9 in both German and English language. Recordings exist of 12 distinct speakers two of which are only present in the test set. The SSC is based on the Speech Commands release by Google which consists of utterances recorded from a larger number of speakers under less controlled conditions. It contains 35 word categories from a larger number of speakers.

Data format

For maximum compatibility, the SHD datasets are provided in HDF5 format which can be read by most major programming languages.


Each datum consists of two lists that contain the firing times and the unit id of which neuron has fired at the corresponding firing time.

Example code

For a tutorial how to train a spiking neural network on this dataset checkout:

The following code illustrates howto download and access the dataset in Python. The example code uses the PyTables package (https://www.pytables.org) to load HDF5 files.

import os
import urllib.request
import gzip, shutil
from tensorflow.keras.utils import get_file
print("Using cache dir: %s"%cache_dir)
# The remote directory with the data files
base_url = "https://zenkelab.org/datasets"
# Retrieve MD5 hashes from remote
response = urllib.request.urlopen("%s/md5sums.txt"%base_url)
data = response.read() 
lines = data.decode('utf-8').split("\n")
file_hashes = { line.split()[1]:line.split()[0] for line in lines if len(line.split())==2 }
def get_and_gunzip(origin, filename, md5hash=None):
    gz_file_path = get_file(filename, origin, md5_hash=md5hash, cache_dir=cache_dir, cache_subdir=cache_subdir)
    if not os.path.isfile(hdf5_file_path) or os.path.getctime(gz_file_path) > os.path.getctime(hdf5_file_path):
        print("Decompressing %s"%gz_file_path)
        with gzip.open(gz_file_path, 'r') as f_in, open(hdf5_file_path, 'wb') as f_out:
            shutil.copyfileobj(f_in, f_out)
    return hdf5_file_path
# Download the Spiking Heidelberg Digits (SHD) dataset
files = [ "shd_train.h5.gz", 
for fn in files:
    origin = "%s/%s"%(base_url,fn)
    hdf5_file_path = get_and_gunzip(origin, fn, md5hash=file_hashes[fn])
# Similarly, to download the SSC dataset
files = [ "ssc_train.h5.gz", 
for fn in files:
    origin = "%s/%s"%(base_url,fn)
    hdf5_file_path = get_and_gunzip(origin,fn,md5hash=file_hashes[fn])
# At this point we can visualize some of the data
import tables
import numpy as np
fileh = tables.open_file(hdf5_file_path, mode='r')
units = fileh.root.spikes.units
times = fileh.root.spikes.times
labels = fileh.root.labels
# This is how we access spikes and labels
index = 0
print("Times (ms):", times[index])
print("Unit IDs:", units[index])
print("Label:", labels[index])
# A quick raster plot for one of the samples
import matplotlib.pyplot as plt
fig = plt.figure(figsize=(16,4))
idx = np.random.randint(len(times),size=3)
for i,k in enumerate(idx):
    ax = plt.subplot(1,3,i+1)
    ax.scatter(times[k],700-units[k], color="k", alpha=0.33, s=2)
    ax.set_title("Label %i"%labels[k])


Copyright 2019-2021 Benjamin Cramer & Friedemann Zenke

The above code is under the MIT License.
The datasets are released under the Creative Commons Attribution 4.0 International License.