cluster_tools

cluster_tools copied to clipboard

The conversion to `uint8` for float data in https://github.com/constantinpape/cluster_tools/blob/master/cluster_tools/copy_volume/copy_volume.py#L208-L214 is incorrect since it just normalizes based on the local statistics, which will lead to artifacts. Instead we need to compute...

constantinpape

the edge feature computations are not working quite correctly: - [ ] affinity features: none of the feature values agree for all edges - [ ] boundary features: the edge...

constantinpape

Support for cloud-based datastores?

4

This looks like a super powerful tool, looking forward to using it! I'd love to implement an API abstraction for cloud datastores like [BossDB](https://bossdb.org) or [CloudVolume](https://github.com/seung-lab/cloud-volume) so that one could,...

j6k4m8

Replace kernighan lin [here](https://github.com/constantinpape/cluster_tools/blob/master/cluster_tools/multicut/solve_global.py#L44). Make sure that the corresponding nifty version is released beforehand.

constantinpape

constantinpape

Axon-Dendrite Attribution code

2

Hi, This file is currently empty: cluster_tools/publications/leveraging_domain_knowledge/1_axon_dendrite_attribution/1_multicut.py it is possible to get this code and demo data? Matthew

MatthewBM

Not all links are there yet: https://github.com/constantinpape/cluster_tools/blob/master/publications/leveraging_domain_knowledge/README.md#obtaining-the-data

constantinpape

This should be not too hard to fix, by exposing and using the `maxNodeId` of `nifty.distributed.Graph`.

constantinpape

Because graph computation and feature extraction expect n5 input data. To fix this, we would need to support hdf5 inputs for these computations in nifty.

constantinpape

support slicing stacks in arbitrary dimension

3

Hi, this should enable extracting only sub-dimensional parts of stacks. (-> single channels of multi-channel stacks for MoBIE etc...) I have not checked how it can be tested yet.

martinschorb