0.4.1

What's Changed

• fix best_metric < worst_metric by @FynnBe in #50
• refactor _train_epoch and _validate by @FynnBe in #51
• refactor DefaultTrainer.from_checkpoint by @FynnBe in #53
• Deserialize ROI Wrapper WIP by @constantinpape in #55
• fix error connected to retain_grad() by @JonasHell in #57
• Add Serializer to DefaultTrainer by @FynnBe in #54
• fix typo in augmentations by @JonasHell in #59
• Add sampler that checks if there are a minimum number of instances in the volume by @JonasHell in #60
• ensure correct behaviour of filters when set manually by @JonasHell in #61
• Add functionality to mask to-bg-boundaries by @JonasHell in #62
• Implement advanced rf training for s2d by @constantinpape in #63
• Update s2d features, extend mito-enhancer training by @constantinpape in #64
• Enable overwriting the min-shape in modelzoo export by @constantinpape in #65
• Extend shallow2deep experiments by @constantinpape in #67
• S2D experiments for training LM enhancers by @constantinpape in #71
• add noise augmentations for mean teacher approach by @JonasHell in #68
• fix typo/bug in validation logging by @JonasHell in #73
• Add axon myelin training by @constantinpape in #74
• add worst_tiles sampling by @JonasHell in #77
• Extend s2d functionality by @constantinpape in #75
• Elastic 3d by @Buglakova in #79
• adapt default mean teacher augs to work with modelzoo export by @JonasHell in #82
• adapt _serialize_feature_config() to work with strings by @JonasHell in #83
• Add tests for augmentations by @constantinpape in #80
• Add distance transform to label transforms by @constantinpape in #84
• Add save_root argument to trainer to enable saving files at different… by @constantinpape in #85
• Use ApplyMask in other masking loss wrappers by @constantinpape in #86
• Make bioimageio import optional by @constantinpape in #88
• fix bioimageio import in shallow2deep by @JonasHell in #89
• Use tifffile to read compressed tif files by @Czaki in #93
• When loading data for train allow to read data that has channel on first position by @Czaki in #94
• Implement sampler for the image collection dataset by @constantinpape in #96
• provide reasonable default values for augmentations by @JonasHell in #97
• Add resnet 3d implementation (based on torchvision) by @constantinpape in #99
• Update livecell.py by @constantinpape in #100
• Augmentations by @JonasHell in #98
• Update spoco losses by @constantinpape in #87
• implement gradient masking with cropping - new PR by @JonasHell in #101
• Add Separate Label Path to the LiveCELL Loader by @anwai98 in #104
• Update Label Type Argument for LiveCELL by @anwai98 in #105
• Raise value error if dice is called with non-matching shapes by @constantinpape in #106
• add args to adjust tqdm output by @JonasHell in #108
• add dense accumulate as additional baseline by @JonasHell in #107
• Add neurips cell seg dataset; update raw image collection dataset; up… by @constantinpape in #111
• Update Checksum for UroCell Dataloader by @anwai98 in #109
• Misc changes WIP by @constantinpape in #110

New Contributors

• @Buglakova made their first contribution in #79
• @Czaki made their first contribution in #93
• @anwai98 made their first contribution in #104

Full Changelog: 0.4.0...0.4.1