Skip to content



A dataset, can consist of multiple layers, which may be

  • segmentation layers or
  • color layers.

Layers contain image data in one or multiple magnifications or short mags (see mipmap or image pyramids for similar concepts). The magnification 4 or 4-4-4 describes a downsampling factor of 4 in each dimension, 4-4-2 specifies anisotropic downsampling. The image data in full resolution is referred to as the finest mag, e.g. 1-1-1, downsampled variants are more coarse.

The voxel size describes the size of a voxel in mag 1, the default unit is nm if not specified otherwise.

The underlying file formats zarr and wkw use chunks as a compressible unit and shards as a storage unit. A wkw block corresponds to a chunk, and a file to a shard.


An annotation can consist of one or multiple annotation layers, which can either be a

  • volume annotation layer, or volume layer in short (main guide), or
  • skeleton annotation layer, or skeleton layer and skeleton in short (main guide).

Volume-only and skeleton-only annotations are restricted to the specific annotation layer type.

A skeleton consists of trees with nodes and edges, which may be organized in groups. A tree is supposed to be a single connected component. Despite the name, it may contain cycles.

Branchpoints are boolean markers on nodes, that can be jumped to easily in the frontend. They are not tied to any network-specific properties of the node.

Tasks & Projects

A task specifies a desired annotation result. A task instance refers to the concrete copy of a task assigned to a user for annotation work. A task may be handed out to several users for redundancy resulting in several task instances, each containing their respective annotation. Task instances may also be abbreviated as tasks if the context is clear.

A project is a collection of tasks.

See also the task and projects guide.

Back to top