Skip to content

v0.3.0
Compare
Choose a tag to compare
@github-actions github-actions released this 07 Aug 20:25
· 169 commits to main since this release
v0.3.0
5e65751

v0.3.0 (2024-08-07)

Chore

  • chore(deps): bump scikit-learn from 1.5.0 to 1.5.1

Bumps scikit-learn from 1.5.0 to 1.5.1.

updated-dependencies:

  • dependency-name: scikit-learn
    dependency-type: direct:production
    update-type: version-update:semver-patch
    ...

Signed-off-by: dependabot[bot] <[email protected]> (95001d5)

  • chore(coco_utils): avoid circular import for tests (551fe22)

  • chore: update npm packages (b8b245f)

  • chore(deps-dev): bump vite from 4.5.2 to 4.5.3 in /vue-components

Bumps vite from 4.5.2 to 4.5.3.

updated-dependencies:

  • dependency-name: vite
    dependency-type: direct:development
    ...

Signed-off-by: dependabot[bot] <[email protected]> (e57e3b7)

  • chore(deps): bump scikit-learn from 1.4.2 to 1.5.0

Bumps scikit-learn from 1.4.2 to 1.5.0.

updated-dependencies:

  • dependency-name: scikit-learn
    dependency-type: direct:production
    update-type: version-update:semver-minor
    ...

Signed-off-by: dependabot[bot] <[email protected]> (3099684)

Ci

Feature

  • feat(image_list): add visible columns control (1c87a6e)

  • feat(transforms): async display of ground truth annotations (6b434ee)

  • feat: use kwcoco for coco datasets (782d6a3)

  • feat(dataset): add force_reload parameter (cfb2b63)

  • feat(transforms): always show ground truth

Always compute score against ground truth. (5c6cde6)

  • feat(image_list): replace grid switch with toggle button (b95031d)

  • feat(ImageDetection): only show annotation ID for ground truth

Was showing repetitive category ID for object detection model
annotations. (396e3e4)

  • feat: toggle ground truth/predictions annotations
  • Toggle ground truth/predictions annotations for source and
    transformation images.
  • Compute COCO score (NRTK) instead of embeddings cartesian distance.
  • Add unit tests for the coco_utils.py newly module. (0ab0b5b)

  • feat(image_list): add search and fullscreen controls (3196272)

  • feat(ImageDetection): keep tooltip visible and hide scrollbars

  • fix img css position to block to avoid following image
    content from shifting as image did not contribute to component width.

  • refactor canvas context creation. (d235d14)

  • feat(image_list): add embeddings distance to image table (1d4f6dd)

  • feat(image_list): add table panel under tab (172b256)

  • feat(image_list): spinner for loading images (1c162e0)

  • feat(image_list): group original and transformed images (6f0d803)

Fix

  • fix(core): filter image selection by ground truth categories

Closes #74 (d0ccc0a)

  • fix(embeddings): hover on original image highlights (0b7d00b)

  • fix(dataset): prefer implicit return types

get_dataset() returns a CocoDataset, not a Dataset. (bc5b0ca)

  • fix: entry name of transforms subapp (363716d)

  • fix(ImageDetection): keep the same color for same labels (9e0a733)

  • fix: dataset.py snake case and remove unused imports (53b6005)

  • fix: show object detection label if no category in dataset (b4f9b41)

  • fix(transforms): compute score for truth to transform correctly (141bea6)

  • fix: score 0 when no predictions, dataset switching bug (9cb8ab1)

  • fix(ImageDetection): keep tooltip from clipping under table

Shifting tooltip if out of window was not enough
when table size was small, and tooltip would clip
under the table footer. (a670d92)

  • fix(transforms): resize transformed image for scoring

Annotation similarity scoring requires the transformed image
to be the same size as the original image. (86c0e2e)

  • fix(transforms): use class agnostic scorer

Avoids error when Object Detection Model outputs
category that is not in COCO JSON. (d3bccfe)

  • fix(image_meta): add distance to meta type dict (43a704e)

  • fix(pyproject): set min versions for nrtk and timm (01ce0bb)

  • fix(ScatterPlot): after selection, start navigation mode again (eb65f38)

  • fix(embeddings): allow selection of transformed point (dd1650f)

  • fix(ImageDetection): avoid extra bottom padding on hover border (95235a5)

  • fix(ImageDetection): miss aligned annotations for transformed images

When transformed image resolution was different from the original image,
the annotations were not aligned with the image.

closes #60 (a38e30f)

  • fix(embeddings): distinguish image kind on hover (26a670d)

  • fix(image-list): show hover box on original or transformed (e61e66c)

Performance

  • perf(transforms): only create transformed images if visible (e11b422)

Refactor

  • refactor(trame_utils): simplify and doc change_checker (0ac0975)

  • refactor(image-list): add image server

Keeps original image off state

Remove commented code (16a953d)

  • refactor(transforms): dont duplicate annotations on context

Stashing them on state once is enough (65e5c8f)

  • refactor(transforms): put state.loading_images in _update_images (2f3e244)

  • refactor(object_detector): add return type annotation (2237edf)

  • refactor(image_ids): remove domain types (834a664)

  • refactor(object_detector): eval returns image_id keyed dict (041a6eb)

  • refactor: fix typing on object_detector, nrtk_transforms

and trame_utils (ffe0349)

  • refactor(transforms): remove pointless state.image_kinds (2308d64)

  • refactor: add image_ids and trame_utils modules (07b254c)

  • refactor(image_list): use Quasar table grid mode

instead of custom grid view. (3fd7629)

  • refactor(transforms): index image meta by dataset ID (05fd731)

  • refactor(embeddings): don't short circuit if id is empty string (6696afe)

  • refactor(transforms): wrap sync _update_images in async _set_source_images (a7128d3)

  • refactor(transforms): rename image_kinds.image_id_key to image_ids_list (20af742)

  • refactor(image_list): use Getter for image_id (6460f64)

Unknown

  • Merge main to release (d67665e)

  • embeddings,object_detector: add OOM recover fix (7cfba68)

  • Auto-merge release back to main (983afc9)

Assets 4
Loading