Rosie Wood
Rosie Wood
https://github.com/openjournals/joss-reviews/issues/6434#issuecomment-2041626766 Fix typos mentioned here
As we are updating our dependencies and supported python versions it would be good to address the large number of deprecation warnings that are produced when running tests. As part...
Worth also to look at ICDAR map text competition output structures https://rrc.cvc.uab.es/?ch=28 https://knowledge-computing.github.io/mapkurator-doc/#/docs/modules/spot (look for training data here)
**Describe the bug** Would be useful to review and streamline dependancies to ensure easy installation but still efficient/effective code. To do: - [ ] Make a list of dependencies and...
**Is your documentation update related to a problem? Please describe.** Some big changes have been made to the code since v0.3.3. We should have a file in the docs which...
At the moment is very difficult to identify training params etc. when looking at model outputs. At minimum we should save these in the `info.txt` file that is created when...
**Is your feature request related to a problem? Please describe.** At the moment, we save: - object checkpoint (i.e. `ClassifierContainer` object) - model checkpoint (i.e. whole model) - model state_dict...
Currently the only post-processing option is to filter for patches with no surrounding patches labelled as the same thing (e.g. an isolated railspace patch) and relabel these as non-railspace if...
### Summary The current version of MapReader contains unused/untested code for using the patch context when training the model. From the 2021 paper: "**Context-aware patchwork method**. A limitation of the...