Data repository

A collection of resources for high-resolution connectomics

Browse the repository online here

How to mount the data repository

The data repository can be found here or mounted as descibed below.

There is no password or username necessary for read-only access.

In Windows:

Simply press “Map network drive” from the Windows Explorer and enter the link provided above.

In Linux:

sudo mount -t davfs SOME_MOUNTPOINT (make sure you have the related packages)

Brief explanantion of content

The content is located in two main folders:

  • Classifier, the trained SynEM classifiers including the interface annotations in the training and validation regions as well as the precalculated interface features
  • TestSet, the raw and segmentation data of the test set and ground truth annotations for all synapses and excitatory spine synapses
  • The data used to train and evaluate SynEM on a dataset from ATUM-SEM (Kasthuri et al., 2015)

Detailed explanation of content

Supplemental Material

Classifier (/Classifier)

  1. The boosted decision-stump classifier trained on all features (SynEMPaperClassifier.mat) and only on the features with positive feature importance (SynEMPaperClassifierPred.mat)
  2. The 40 training/validation volumes containing interfaces and interface labels (/Classifier/SynapseDetectionTrainingDataRevised)
  3. The features calculated from the training and validation cubes (/Classifier/train, /Classifier/val)

Test set (/TestSet)

  1. The SynEM test set ground truth containing all synapses (TestSet_allSyn_gt.mat) and excitatory spine synapses only (TestSet_excSyn_gt.mat)
  2. The SynEM test set raw data and segmentation (TestSet_raw_seg.mat) (see also +SynEM/+Examples/synEMWorkflow.m in the SynEM repository)

Ground truth data and the trained classifier used for the evaluation of SynEM on the ATUM-SEM dataset from Kasthuri et al. (2015). See also the scripts downloadDatasetScript.m, labelGeneration.m and trainSynEM.m in the KasthuriData package in the SynEM repository.

  1. The modified segmentation created for the evaluation of SynEM as described in the “Evaluation on the dataset from Kasthuri et al. (2015)” method section in the SynEM paper.
  2. KasthuriDirClassifier.mat: The trained SynEM classifier.
  3. SynapseDirectionLabels_v3.mat and synapseDirectionGT_v3.mat: The interface ground truth data.
  4. local/: Tiling of the dataset into non-overlapping cubes containing the supervoxel graph data for each local cube and the SynEM interface features.
  5. global*.mat files: The supervoxel graph data collected from the single tiles in local/*.
  6. allParameter.mat: The paths to the folders and files in local/* (see p.local).