v0.5.0-beta

What's Changed

• Add background class for segmentation if pixels without annotations found by @sokovninn in #191
• Clip width and height in BBoxAnnotation by @klemen1999 in #192
• Updated auto labeler configuration by @kozlov721 in #195
• Fixed records without annotations by @kozlov721 in #197
• Update docs generation by @kozlov721 in #198
• LuxonisML Data Docs by @kozlov721 in #196
• Fix: Prevent training from crashing when unable to connect to MLflow. by @JSabadin in #193
• fix: sort classes alphabetically by @JSabadin in #194
• fewer reconnections by @JSabadin in #203
• hotifx: adding file names to be ignored by the COCO parser by @JSabadin in #199
• fix: add background to multiclass only by @JSabadin in #204
• Fix augmented loader with no labels by @kozlov721 in #202
• Added pydoctor to CONTRIBUTING.md by @kozlov721 in #201
• Speed up semantic segmentation annotations loading by @sokovninn in #205
• Added dai_type to NNArchive by @klemen1999 in #200
• Release v0.5.0 by @klemen1999 in #206
• docs: fixed warnings and improved formatting by @kozlov721 in #207

Full Changelog: v0.4.1-beta...v0.5.0-beta

v0.4.1-beta

What's Changed

• [Hotfix] Config version set to 1.0 by @klemen1999 in #190

Full Changelog: v0.4.0-beta...v0.4.1-beta

v0.4.0-beta

What's Changed

• Layout Inference by @kozlov721 in #153
• Updated fsspec Version by @kozlov721 in #155
• Small Typing and Code Fixes by @kozlov721 in #156
• Added __iter__ to BaseLoader by @kozlov721 in #157
• DevOps Updates by @kozlov721 in #158
• Augmentation SpeedUp fix by @sokovninn in #154
• Fixed path retrieval in loader which caused slow training on GCP by @JSabadin in #159
• Fixed CLI Table Width by @kozlov721 in #161
• kwargs in setup_logging by @kozlov721 in #162
• Removed Singletone Pattern from LuxonisConfig by @kozlov721 in #160
• Coverage 90% by @kozlov721 in #163
• Fix Loading Annotations by @kozlov721 in #164
• Updated CODEOWNERS by @kozlov721 in #165
• Augmentations hot-fix by @JSabadin in #167
• Removed rich.print For Debugging by @kozlov721 in #170
• Simplified make_splits Parameters by @kozlov721 in #169
• Duplicate Data Warning by @JSabadin in #168
• Fixed Order of Transforms in Augmentations by @klemen1999 in #171
• Added typeguard to requirements-utils by @kozlov721 in #172
• Fixed guard_extras to raise ImportError by @kozlov721 in #173
• Configured codecov by @kozlov721 in #174
• Removed numpy.typing by @kozlov721 in #175
• v0.4.0 - Metadata Update by @kozlov721 in #178
• [Feat] Added name attribute to HeadType by @klemen1999 in #176
• Added automatic clipping for normalized annotations by @klemen1999 in #177
• Formatting by @kozlov721 in #179
• Speedup in picking batched aug indices by @klemen1999 in #183
• Fixed Incorrect Pyproject Classifier by @kozlov721 in #185
• Improved Duplicate Warnings by @kozlov721 in #184
• Bug Fix: Filtering Issue & Mosaic Transformation Logic Improvement by @JSabadin in #186
• Embedded Python Badge by @kozlov721 in #187
• LuxonisParser - RoboFlow URL Support by @kozlov721 in #189
• LuxonisML - v0.4.0 by @kozlov721 in #188

Full Changelog: v0.3.0-beta...v0.4.0-beta

v0.3.0-beta

What's Changed

• luxonis_ml.data
  • Suport for custom tasks when using parsers (#141)
  • Support for using pixel mask annotations for generating LuxonisDataset (#142)
  • COCO Parser now supports both RoboFlow and FiftyOne COCO formats (#137)
• luxonis_ml.tracker
  • Support for native MLFlow system metric monitoring (#146)
  • Easier artifact uploading using LuxonisTracker.upload_artifact method (#143)

Bugfixes

• Fixed freezing LuxonisFileSystem.walk_dir with GCS protocol (#148)
• Fixed various issues related to incorrect syncing of data when using remote storage for LuxonisDataset (#128)
• Several performance improvements
  • Augmentations speedup (#147)
  • LuxonisLoader speedup (#140)
• RLE counts can be zero (#144)

v0.2.3-alpha

What's Changed

• Fix - NN Archive Validation by @kozlov721 in #123
• Correctly pass number of segmentation classes to augmentations by @klemen1999 in #131
• Fix and Enhance Head Metadata Validation by @jkbmrz in #132

Full Changelog: v0.2.2-alpha...v0.2.3-alpha

v0.2.2-alpha

What's Changed

• Parsers in NN Archive #127
• Task ingestion #126

Breaking Changes

• NN Archive structure changed

v0.2.1-alpha

Bugfixes

• #123

v0.2.0-alpha

What's Changed

• Speed optimizations
  • ~10 times faster data processing
  • #117
• Code coverage
  • From 66 % to 82 %
• Various NN Archive improvements
  • #106
  • #109
  • #120
  • #112
• Annotations refactor
  • #110
  • Support for custom tasks
    • allows using multiple annotations of the same type in a single dataset
  • clearer structure of annotation dictionaries

Breaking Changes

Compatibility with datasets created using previous versions?
Not compatible. All datasets must be re-created.

• Refactored annotations
  • Changed structure of dictionaries yielded from the dataset generator
  • See #110
• Changed backend for data handling
  • migrated from pandas to polars
  • in case of errors:
    • delete luxonis_ml directory located at $LUXONISML_BASE_PATH/luxonis_ml
      • $LUXONISML_BASE_PATH is set to your home directory by default