Skip to content

KNOSSOS 5

Compare
Choose a tag to compare
@my-tien my-tien released this 16 Jan 00:05
· 1244 commits to master since this release

Features

Dataset:

The dataset loading mechanism was updated to support a wider range of dataset configurations, including

  • local jpeg-compressed raw datasets (additionally to remote ones)
  • customizable dataset cube sizes
  • various dataset backends. For example a dataset url containing “/ocp/ca/” will be interpreted as an open connectome dataset.

Mesh:

It is now possible to load triangle meshes into KNOSSOS. Meshes are a set of vertices, colors and indices. In KNOSSOS they are associated to one skeleton tree and are displayed inside the 3D viewport.

Simply insert a mesh file (.ply) into a KNOSSOS annotation file (.k.zip) and it will be loaded with it.

Visit our wiki to read about the exact file format.

Inside KNOSSOS you can jump to specific areas of interest: Just click on the mesh location and press S.

mesh_screenshot

Skeletonization:

  • Trees are now traversed along the edges (depth-first-search) with X and Shift+X (previously, traversal order was defined by node ID).
  • You can now enable anti-aliasing for a smoother skeleton visualization.
  • Node comments can be displayed directly inside viewports. This should improve your orientation.

Synapses feature:

For neural annotations you can now create synapses with a synaptic cleft as
well as pre- and postsynapse:

  1. Simply perform your ordinary skeletonization.
  2. On encountering a synapse, end the current tree with Shift+C. The last placed node is marked as pre-synapse.
  3. Now annotate the synaptic cleft by placing nodes along it as usual. When you are finished, press C to end the cleft.
  4. The next placed node will be marked as post-synapse.

screenshot from 2016-12-31 16-04-10

Properties feature:

Until now additional information about trees and nodes had to be crammed into the respective object’s comment section. This became unreadable and hard to parse very quickly.

So we introduced customizable and visualizable properties. Properties are either simple text or textual or numerical key value pairs.

Numerical node properties can be highlighted inside the viewports: You can choose to scale radius by property value or to map property value to color map value, or both.

Segmentation:

  • You can now perform bucket fills, with the viewport being its maximum scope (middle mouse button)
  • Segmentation operations are now possible in all magnifications instead of only in mag 1.
  • The form of the segmentation brush is now aligned with the image pixels.
  • The color of segmentation objects can be changed.

Annotation in general

  • You can use the dropdown in the toolbar to quickly switch between different
    work modes, e.g.:
  • “Tracing Advanced“: More complex tracing operations such as unlinked
    nodes or multiple trees have been moved into this separate work mode,
    so that ordinary tracing has become simpler and less error-prone.
  • “Merge Tracing”: In this mode skeletonization and segmentation are
    performed simultaneously: Place nodes on segmentation sub-objects to
    automatically merge them together. The result will be a classical
    skeletonization in which each tree corresponds to one merged
    segmentation object.
  • “Review”: A segmentation proof-reading mode in which painting and
    merging are disabled to prevent accidental modifications.
    Skeletonization serves as review annotation.
  • A movement area can be defined to restrict your work to a specific location
    of interest. Everything outside this area is darkened to an adjustable
    degree and cannot be modified.
  • Spacebar hide: The viewport is often cluttered with nodes, edges,
    segmentation overlay and viewport tools. Hold the spacebar and everything
    except the raw dataset will be hidden for this duration.

Python Plugin Manager:

KNOSSOS now has a plugin manager, allowing you to write and share plugins written in Python in an online repository.

For example you can write custom GUI elements, create annotations programmatically or utilize your own algorithms. The idea is to customize KNOSSOS and make it fit your unique workflows.

As a starting point and reference we have a set of plugins available at https://github.com/knossos-project/knossos-plugins.

The Manager can be found under the “Scripting” menu entry.

Viewports:

  • You can now take viewport snapshots of varying resolutions. The snapshots can be configured to decide what should be visible in them, e.g. only raw data, segmentation, skeletonization etc.
  • Viewports can be undocked and positioned independently of the KNOSSOS window. Especially, multi-screen workplaces might find it useful to place a fullscreen viewport onto one screen.
  • A scalebar in every viewport shows the dataset’s physical dimensions.
  • The 3D viewport now supports wiggle stereoscopy, just hold W. Wiggle stereoscopy is a computer graphics method to make the 3-D nature of structures better recognizable.
  • Inside the plane viewports KNOSSOS now shows information about what’s underneath your mouse cursor:
    • The position inside the dataset (shown in the lower left corner)
    • The segmentation object id at that position (shown below the segmentation table in the annotation widget)
  • A fourth viewport with arbitrary orientation can be turned on. One of the purposes of an arbitrary view is to automatically orient itself orthogonally to the skeleton while tracing. This feature is still experimental.

User Interface:

  • In the Annotation widget we replaced the previous skeleton view that caused performance problems for large skeletons. The new table also supports sorting.
  • Skeleton as well as segmentation tables consume less space. There is only
    one segmentation table, but if overlapping objects are selected, they are
    shown in an extra table.
  • We moved various options into preferences widget to reduce the number of
    floating windows in KNOSSOS.
  • When loading a dataset, you can specify a field of view for the viewports.
  • We tried to make all shortcuts visible in the user interface.
  • Comment shortcuts have moved from the F keys to the number keys.

Zoom:

  • The zoom interface now supports dynamic switching between magnifications.
    A value of 100% represent the highest magnification.
  • In the 3D viewport KNOSSOS now zooms in at the cursor position. To inspect
    something closer, simply let the mouse cursor hover over it while zooming.
  • In conformance with most other user interfaces, Ctrl++, Ctrl+- and
    Ctrl+0 are now the shortcuts for zooming in, out or to reset zoom.

Other improvements

  • KNOSSOS does not consume as much computing power anymore while running in the
    background.