Robert Sachunsky
Robert Sachunsky
IIUC, in the current implementation, the rate between original images and augmented variants will be a fixed `1 : N` for `N` kinds of augmentation that have been enabled: https://github.com/qurator-spk/sbb_pixelwise_segmentation/blob/63fcb961898540f95cb347d43f8965ae65b8be3f/utils.py#L92...
The current implementation of the rotation augmenter does not actually rotate or reflect, it transposes: https://github.com/qurator-spk/sbb_pixelwise_segmentation/blob/63fcb961898540f95cb347d43f8965ae65b8be3f/utils.py#L108-L113 Thus, the image will still be flipped / mirrored afterwards. I suggest using an...
| **original** | **on #173** | **on #175**| | --- | --- | ---| |  |  |  | |  |  |  | | ...
- `_page_worker`: remove `ThreadPool` mechanism introduced in 3cc47804 (which broke processors that are not threadsafe like TF/Keras) - since _no mechanisms_ work to stop computation in uniprocessing (as not even...
(Posting this in German to re-use a chat...) `OcrdEnvConfig` ist sehr seltsam definiert: Nehmen wir eine Variable, die keinen Default hat, z.B. `OCRD_METS_CACHING`: Der Versuch einer Referenzierung (`config.OCRD_METS_CACHING`) ergibt eine...
builds upon both #465 and #449, merging both conceptually idea is to still rely on Makefile as much as possible, allow user to switch between client (ocrd_network) and standalone (local...
Since we agreed it is not feasible to continue supporting native installation into a shared venv (along with risk of dependency clashes, necessity to find compromises or sidestep via sub-venv),...
In https://github.com/qurator-spk/ocrd_repair_inconsistencies, what our PageValidator reports as [ConsistencyError](https://ocr-d.de/core/api/ocrd_validators/ocrd_validators.page_validator.html#ocrd_validators.page_validator.ConsistencyError) gets repaired by trying to re-order via coordinate centroids (in tandem with `@textLineOrder` for regions and `@readingDirection` for lines). Whether or not...