Releases: comet-ml/kangas
Releases · comet-ml/kangas
2.3.0
New in this release:
- New
kangas.Tensor()
for vectors to store in asset database (as opposed to just lists (eg, "vectors") that are stored in the database directly - Added
Computed Columns
, available in the top row controls as+
- You can group, sort, and filter on computed columns
- Embedded points selected are save in clipboard for use in filters and computed columns
- New examples on kangas.comet.com: embeddings, and breast_cancer
2.2.10
New in this release:
- Updated kg.Embedding() support for "pca" and "t-sne" projections
- kg.Embedding() also can set text, name, include, and color in constructor
- New cache makes projections faster
- Updated embedding thumbnail drawings to respect global x/y ranges
2.2.9
2.2.7
2.2.6
2.2.5
2.2.2
New in this version:
- Label colors may have changed for those labels with more than 5 characters; the previous algorithm was JavaScript-specific. Now, it can replicated in other languages
- Added Image.add_mask() and Image.add_mask_metric()
- Thumbnail images now show annotations, including masks, bounding boxes, regions, and lines
- In the full image views, regions now show a border
- Vector data now properly shows as JSON
- Column names cannot contain ":" or ","
- Fixed a bug in DataGrid.add_iou_column()
- Many bug fixes
2.2.1
2.2.0
New in this release:
- Fixed multiple issues in horizontal scrolling: boundary updates, throttle events
- Suspension boundaries are correctly installed
- Fetches now use cache
- Requires flask 2.2 or greater
- Mask class for easily manipulating masks
- Importing kangas now loads much faster
- New download datagrid for hosted services
- Much faster segmentation mask rendering
- JsonViewer now has copy button, and shows element sizes
- Bounding box annotations now show labels
- Labels now show layer name, if given
- Added DataGrid.add_iou_columns() for automatically added Intersection Over Union columns
- Many bug fixes and enhancements
2.1.0
New in this release:
- New annotation types:
Image.add_marker()
- "raindrop" and "circle" typesImage.add_line()
- draw a lineImage.add_mask()
- segmentation-style masks (every pixel is a class id)Image.add_mask_metric()
- heatmap-style mask (every value is a measurement of a metric, 0 - 1)
- Bounding boxes now have labels showing label name and score
- Annotations now have layers
DataGrid.add_iou_columns()
- adds "Intersection Over Union" visualization and measurement columns- New JSON Viewer
- Fixed zoom in expanded image viewer
- New flags for server:
--backend-port
and--backend-proxy
- Many bug fixes