v1.9.1

What's Changed

• load pytorch ckpt files for pretrained vits by @themattinthehatt in #304
• pypi release workflow by @themattinthehatt in #305

Full Changelog: v1.9.0...v1.9.1

v1.9.0

What's Changed

• update colab notebook by @themattinthehatt in #292
• remove unnecessary jupyter notebook data by @themattinthehatt in #293
• remove neck in vits to avoid over stacking issues by @PPWangyc in #275
• fix vit_b_sam predict model init issue by @PPWangyc in #294
• Add a hidden run_app command by @ksikka in #298
• typecheck fix by @themattinthehatt in #300
• implement SAM backbone with huggingface transformers package by @themattinthehatt in #301
• new backbone: ViT pretrained on imagenet with mae loss by @themattinthehatt in #302
• pretrained dino backbone by @themattinthehatt in #303

New Contributors

• @PPWangyc made their first contribution in #275

Full Changelog: v1.8.1...v1.9.0

v1.8.1

What's Changed

• Making sure path is str by @hummuscience in #286
• bug fix in left-right/top-down augmentation pipeline by @themattinthehatt in #287
• update/put optimization defaults in one place by @themattinthehatt in #288

Full Changelog: v1.8.0...v1.8.1

v1.8.0

What's Changed

• CI: run on ubuntu-latest by @ksikka in #253
• CI: fix workflow cancellations by @ksikka in #254
• CI: use tt flag to see if it fixes SIGINT propagation by @ksikka in #255
• repo cleanup by @themattinthehatt in #257
• Fix multigpu test by @ksikka in #260
• imgaug refactor by @themattinthehatt in #259
• Add Multiview predict on videos method by @ksikka in #258
• Two small but important fixes by @ksikka in #265
• clean up sleap2lp converter by @themattinthehatt in #266
• [Testing] Patch keypoint_names with value of keypoints by @ksikka in #271
• Add flag to CLI to skip already predicted videos by @hummuscience in #269
• [Needs docs] Add optimizer to config by @ksikka in #270
• Refactor: split cli/main.py into one file per command, making it easier to add commands by @ksikka in #273
• Fixup cli prediction code post AI-driven refactor by @ksikka in #276
• head refactoring by @themattinthehatt in #279
• Replace : with - in default model directory by @ksikka in #280
• Migrate to pyproject.toml and update project structure by @ksikka in #281
• docs refactoring by @themattinthehatt in #285

Full Changelog: v1.7.1...v1.8.0

v1.7.1

Main contributions are functioning CI and the cropzoom pipeline (see here)

What's Changed

• Minor fixes by @ksikka in #241
• Automatically set num_workers to cpu count by @ksikka in #243
• Add option to use steps instead of epochs by @ksikka in #242
• add NaNPCA transform method and tests by @themattinthehatt in #245
• Update run-tests.yml by @ksikka in #246
• Skip failing cropzoom test for now by @ksikka in #251
• Simplify metric code by @ksikka in #247
• Add CI scripts and documentation by @ksikka in #249
• CI: cancel previous build on push by @ksikka in #252
• Add support for cropzoom pipeline by @ksikka in #250

Full Changelog: v1.7.0...v1.7.1

v1.7.0

What's Changed

• make cropzoom tests compatible with python 3.8 by @themattinthehatt in #229
• get rid of unnecessary lightning logs during inference by @themattinthehatt in #230
• Fix issues reading labels file with empty first row by @ksikka in #234
• Split up train, add Model class, add litpose CLI by @ksikka in #232
• Add docs on Model, litpose by @ksikka in #237
• Upgrade some union, list, dict annotations to 3.10 by @ksikka in #239

Full Changelog: v1.6.1...v1.7.0

v1.6.1

What's Changed

• Save copy of labeled data csv to output dir by @ksikka in #222
• Fix tests and post-training analysis behavior by @ksikka in #221
• Save out detector artifacts on detector training by @ksikka in #224
• fix bad import by @themattinthehatt in #226
• add left-right flip option to augmentations by @themattinthehatt in #227
• update version by @themattinthehatt in #228

Full Changelog: v1.6.0...v1.6.1

v1.6.0

What's Changed

Major updates for this release include multi-GPU training, new imgaug transforms, and many small usability improvements

• update sleap2lp.py by @themattinthehatt in #196
• Upgrade sphinx-rtd-theme to fix search #200 by @ksikka in #201
• Fix search attempt 2. Fix #200 by @ksikka in #202
• Log hydra job config to tensorboard by @ksikka in #204
• Change backbone finetuning strategy to allow for DDP by @ksikka in #205
• Support multi-GPU training for supervised training by @ksikka in #206
• Multi-GPU support for unsupervised learning by @ksikka in #207
• Continuous integration using github actions by @ksikka in #208
• add coarse salt/pepper to imgaug transforms by @themattinthehatt in #209
• Added option to choose colormap for labelled videos by @hummuscience in #211
• complete config file documentation by @themattinthehatt in #213
• Add PCA Centering method (experimental) by @ksikka in #214
• Context frames can use any delimiter by @ksikka in #216
• Replace relative paths with absolute paths in demo config by @ksikka in #217
• Fix doc build failure by @ksikka in #218
• Update README.md with downloads/week badge by @ksikka in #219

New Contributors

• @ksikka made their first contribution in #201

Full Changelog: v1.5.1...v1.6.0

v1.5.1

Several small updates, as well as reverting the default training setup to automatically save out the best model according to validation metrics, rather than the last model.

• make sure validation loader always has at least one sample
• allow for early stopping again, as well as saving out model weights every n epochs
• make training fully deterministic
• increase flexibility of LitDaliWrapper to work with custom dali video pipeline functions
• increase flexibility of keypoint correspondence array for true multiview datasets

v.1.5.0

What's Changed

• Passing vfr_check to fn.readers.video by @hummuscience in #179, #180
• remove augmentations for val/test data by @themattinthehatt in #181
• implement pca object that handles NaNs by @themattinthehatt in #184
• support for multiple data streams by @themattinthehatt @farzad-ziaie @Selmaan in #187

Full Changelog: v1.4.0...v1.5.0