Convml-tt

[Full Changelog](https://github.com/convml/convml_tt/compare/v0.10.1...v0.11.0)

*new features*

- Annotated scatterplots now work for `data.dataset.MovingWindowImageTilingDataset`s
[\52](https://github.com/convml/convml_tt/pull/52)

- `utils.get_embeddings` now uses GPU for producing the embeddings if one is
available, drastically speeding up inference time.
[\53](https://github.com/convml/convml_tt/pull/53)

*breaking changes*

- `utils.make_sliding_tile_model_predictions` has been removed in favour of
using `utils.get_embeddings` directly
[\54](https://github.com/convml/convml_tt/pull/54)

[Full Changelog](https://github.com/convml/convml_tt/compare/v0.10.0...v0.10.1)

*bugfixes*

- Fix for xarray bug which arised with call in
`utils.interpretation.rectpred.plot.make_rgb`
[\51](https://github.com/convml/convml_tt/pull/51)

[Full Changelog](https://github.com/convml/convml_tt/compare/v0.9.0...v0.10.0)

*new features*

- Add writer for tensorboard projector visualizer to enable exploration of
tiles in embedding space
[\43](https://github.com/convml/convml_tt/pull/43)

- support for z-order and zero-offset annotations in annotated scatterplot
[\44](https://github.com/convml/convml_tt/pull/44)

- add isomap to available transforms on rectangular domain
[\47](https://github.com/convml/convml_tt/pull/47)

- make it possible to load fastai v1 model using
`TripletTrainerModel.load_from_checkpoint`
[\46](https://github.com/convml/convml_tt/pull/46)

- add `data.dataset.MovingWindowImageTilingDataset` to produce image tile
data from a sliding window across a larger image
[\45](https://github.com/convml/convml_tt/pull/45)

*breaking changes*

- `utils.make_sliding_tile_model_predictions` now takes in
a `data.dataset.MovingWindowImageTilingDataset` rather than an image,
and so a dataset must be created from an image first
[\45](https://github.com/convml/convml_tt/pull/45)

*changed default*

- annotated scatterplot now sets equal x/y-axis
[\44](https://github.com/convml/convml_tt/pull/44)

*maintenance*

- fix for `pytorch-ligtning >= 1.5.0` and exclude `torchvision == 0.10.*`
(can't load `.tgz` files)
[\48](https://github.com/convml/convml_tt/pull/48)

- Add link to Google Colab instructions
[\48](https://github.com/convml/convml_tt/pull/49)

[Full Changelog](https://github.com/convml/convml_tt/compare/v0.8.0...v0.9.0)

*new features*

- Add before and after training logger for producing a dendrogram plot
[\40](https://github.com/convml/convml_tt/pull/40)

- Add "best triplet" option to tile sampling method for dendrogram plot
which uses triplets for which anchor-neighbor are the closest
[\38](https://github.com/convml/convml_tt/pull/38)

- Add rectpred sample plot function [\30](https://github.com/convml/convml_tt/pull/30)

- Enable dendrogram plots for triplets [\35](https://github.com/convml/convml_tt/pull/35)

- Add interactive rect plot for visualising embedding distance to
a selected part of the domain [\34](https://github.com/convml/convml_tt/pull/34)

- Add one-cycle learning rate scheduler
[\32](https://github.com/convml/convml_tt/pull/32)

*changed default*

- Weight decay (L2-regularisation) is now enabled by default (set to
`0.01`) and the learning-rate has been increased to `1.0e-2` to match
the values used when training the fastai v1 based model. Note this
learning-rate was optimised for for a batch-size of `50` using the
`resnet18` architecture. For different batch-size, weight-decay or
architecture a new learning may be needed (use the [pytorch lightning
learning rate
finder](https://pytorch-lightning.readthedocs.io/en/latest/advanced/lr_finder.html))
[\30](https://github.com/convml/convml_tt/pull/30)

*bugfixes*

- Fix for grid-overview plot always showing same set of tiles rather than tiles with indecies provided [\36](https://github.com/convml/convml_tt/pull/36)

[Full Changelog](https://github.com/convml/convml_tt/compare/v0.7.1...v0.8.0)

**Merged pull requests:**

- Add support for anti-aliased backbones [\28](https://github.com/convml/convml_tt/pull/28) ([leifdenby](https://github.com/leifdenby))
- Rectpred refactor [\27](https://github.com/convml/convml_tt/pull/27) ([leifdenby](https://github.com/leifdenby))
- Bugfix for load of fastai v1 trained model [\26](https://github.com/convml/convml_tt/pull/26) ([leifdenby](https://github.com/leifdenby))

Bugfixes and minor improvements

[Full Changelog](https://github.com/convml/convml_tt/compare/v0.7.0...v0.7.1)

**Merged pull requests:**

- Add black code-check to CI [\25](https://github.com/convml/convml_tt/pull/25) ([leifdenby](https://github.com/leifdenby))
- CLI bugfixes [\24](https://github.com/convml/convml_tt/pull/24) ([leifdenby](https://github.com/leifdenby))
- improvements to CLI [\23](https://github.com/convml/convml_tt/pull/23) ([leifdenby](https://github.com/leifdenby))