DeepSlice 1.2.5

add support for labeling bad sections. This can be useful if you have damaged tissue or a few sections of hemibrain data. Rather than exclude these data or have them stuff up the propagation we can instead label them as damaged and have their position propagated from the other intact sections. It only propagates the position and angle but keeps the original width and height, this may still be inaccurate if the section is highly damaged and for this we recommend QuickNII corrections.

DeepSlice 1.2.4

Deprecate save to XML

DeepSlice 1.2.3

Problem with pinned protobuf prevented package properly installing from PyPi (try saying that 10 times quickly)

DeepSlice 1.2

Allows DeepSlice to use tensorflow 2 and more recent versions of Python.
Thanks to @wjguan for fixing a major model loading bug!

DeepSlice 1.1.7

Full Changelog: 1.1.6...1.1.7

DeepSlice 1.1.6

Fixes an issue where paths of images could be accidentally saved when on Windows

DeepSlice 1.1.5

Added lxml to the dependency list as users have been recieving errors when installing via PyPi. We will phase out XML outputs in the future and then remove lxml.

DeepSlice 1.1.4

First release to include a citable doi via zenodo for users who would like to specify the exact version they used.

DeepSlice 1.1.3

In this release we have added the ability to auto detect potential outlier sections and to manually specify heavily damaged sections. This allows DeepSlice to weight these sections less when calculating average angle and section number based position.

DeepSlice 1.1.2

Added the ability to choose to use a secondary model and fixes a bug with propagating angles when a section is outside of the atlas.