February 15th, 2023
almost 2 years ago by Kili Technology
Changelog (v2.129.2)
New
- [Python SDK] We’ve added two new methods:
kili.inferences()
andkili.predictions()
. You can use them to get the list of labels that were created by ML models (inference
orprediction
label types).
Changed
- [Labeling] In video projects, when you click
Submit
on an asset that has incomplete frame-level jobs, the submission process is stopped and a warning message with the index of the first incomplete frame appears. This way, you’ll be able to locate and fix the problem much faster. - [Labeling] We’ve added support for a wider range of TIFF raster drivers. For details, refer to our documentation.
- [Issues & questions] We’ve improved the accessibility of the resolved issues & questions.
- [Analytics] Team performance graphs are now grouped by user id instead of user name. This will help you spot specific users with more precision.
- [Python SDK] When calling methods that run queries (
.assets
,.labels
etc.), you can now request a specific object type to be returned. Set theas_generator
argument toTrue
if you want the method to return a generator. Ifas_generator
is set toFalse
, the method returns a list. This will give you more flexibility when handling returned data.
Fixed
- [Labeling] The Interactive Segmentation tool used in semantic object detection was not working with PNG images. We’ve fixed this issue.
- [Labeling] When a rotated bbox added to a GeoTIFF image was being resized, the shape of the bbox would sometimes change to shapes other than rectangle. We’ve fixed this issue.
- [Labeling] In text projects with many overlapping annotations, some tags were not being displayed. We’ve fixed this issue.
- [Python SDK] Mutation methods would sometimes end before the mutation was complete, for example: entities (assets or projects) were not available yet on the server. This potentially lead to errors when trying to process these entities. We’ve fixed this issue.