- Correct vits assignment
(
d35ae96)
- Handle index out of range on vits assign
(
2aee31a)
- Rename vits prediction option to --use-vits
(
a1e1301)
-
Correct order of file/byte for running vss (
1d896ee) -
Correct handling of remove corner (
60df1ae) -
Assign exemplar to crop (
05b97c9)
- Updated poetry
(
2d97ed5)
- Switch projects
(
22090f3)
-
Added support for assigning predictions to clusters (
1b674fe) -
Assign unknowns via vss (
947a8a3) -
Added pass through of vss server and renaming cluster id in exemplar output (
67c4202)
- Assign everything, not just clusters and assign top prediction if there is more than a .05 spread
in the top 2
(
f6cf171)
- Always run saliency on multiproc regardless of cpu or gpu as it is not gpu enabled
(
b7e913e)
- Correct boolen for remove corners
(
4161e62)
- Default to include all detections in cluster
(
0bf446a)
- Add yolov5 to build
(
b8c2eba)
- Added --model MBARI/yolov5x6-uavs-oneclass to detection
(
5fcd915)
- Correct CUDA HDBSCAN fit
(
c6a8db3)
-
Disable release (
cbf6394) -
Clean up poetry build before docker (
c74efc8) -
Revert to only only amd cuda build (
dd133f0) -
Revert to only only amd cuda build (
f181dfc) -
Slimmer docker add optional imports (
48ae61a)
- Add to cluster command option --skip-visualization since this takes some time and is not needed
for prod ml workflows
(
48d1f19)
- Minor fix to correct debug statement and correct order arg for cluster
(
645e979)
- Move last cluster removal to exemplars only
(
96fc5db)
- Do not load the last cluster in exemplars
(
bbc3a9a)
- Trigger release with changes to init.py
(
e275d9c)
- Trigger release with changes to pyproject.toml to update version
(
6b7f641)
- Revert changes to author list
(
b7cd5a5)
- Better handling of testing log write permissions
(
728b145)
- Adjust test path
(
83119fb)
- Moved tests to root of project
(
ccf2601)
- Added clustering table, transformer paper, and other minor revisions
(
56b4e17)
- Added latest 30 and 18k models
(
673a7f6)
- Removed color image normalization in preprocessing for SAHI and extract min_std and block_size for
more flexible application across projects
(
d9d954a)
- Pass through device correctly for clustering
(
0996b8c)
- Install cuda enabled torch
(
c01259b)
- Changing 2 parameters to ISIIS images
(
417bbf4)
- Skip over HDBSCAN for small cluster
(
cbf1ea3)
- Disable docker build until resolve issue with out of space
(
59e160f)
- Handle small cluster dataset reduction plot by switching to PCA and reduce build to only linux
(
16ab4de)
- Correct handling of single cluster
(
f6ede19)
- Switch to entrypoint
(
02d4eb2)
- Minor update to intro
(
ec75265)
- Added support for arg --cluster-selection-method eof/leaf to support small clusters which work
better with eof
(
6480ac2)
- Migrated to transformers library with batch size of 8, moved some imports to only where needed for
some speed-up, and removed unused activation maps.
(
c5fe725)
- Handle bad image crops (zero length). Fixes #13.
(
c02394c)
- Switch to PyPi release and skip over large build
(
25d3da3)
- Fixes a bug in handling zero clusters which occasionally happes for very small roi collections.
Closes #12.
(
0f69805)
- Slim with two stage docker build, added explicit semantic release steps for clarity, and test pypi
for pip install
(
ff2a46c)
- Bump version to 1.8.0 still failing in ci
(
ed405e8)
- Add semantic release action to write to repo
(
66be480)
- Slim docker images a tad, switch to generic docker user, and replace with miniconda install for
better torch/cuda support
(
782848d)
-
Simplified docker build within pypi size limits with pip build and renamed accidentally commited backup file (
cfcb581) -
Minor change to more specific CUDA docker version in docker tag (
dd2286b) -
Updated poetry lock file with pytest (
a8b62b0) -
Add pip package called sdcat with poetry build and other minor corrections in comments (
dbe494a)
- Auto default logs for different use cases: testing, local dev, and to /tmp on permission failure
(
0c57d45)
-
Removed unused imports and bump torch to python3.11 compatible and (
57c6c2b) -
Fix docker build (
0d7ab83)
- Remove arm cuda build
(
131d3bd)
- Added tsne as an option, defaulted to false as it is generating too many clusters
(
8f974f3)
- Add sys path for convenience
(
d668246)
- Updated with links visible in dockerhub
(
53c91b0)
-
Fix release version from tag (
5d083df) -
Remove pass through of version to dockerhub (
a43cba7) -
Correct order of tags in build and reduce dockerhub short description to < 100 (
86bedcf) -
Correct order of tags in build (
86f100c) -
Consistent image tag (
63165fb) -
Install tsne separately and clean (
17f5a6e) -
Pin build for torch (
3f32cb6) -
Add docker hub description (
9695146) -
Add CPU only build (
8750d94) -
Correct pypi name for MulticoreTSNE (
4d0702c) -
Correct addition of docker user (
07aff58) -
Adding in Docker CUDA12 build semantic release for amd/arm64 (
38aebba)
- Added export of the examplars, handle small clustering input by bypassing tSNE which fails, and
make dino_vits8 the default
(
2306316)
- Conditional import of multicore tsne
(
76ec895)
- Minor change to remove unused import
(
d87b1de)
-
Check for det columns (
ff9a29d) -
Image size in int not float needed for resize (
cf3ebd3) -
Image size in int not float needed for resize (
18fc005) -
Correct PIL image path (
c0da6b6) -
Added image width/height and fixed multiproc to square (
b7e2f21) -
Path to string (
977ef32) -
Roi_dir needs to support lists (
3808bf9) -
Removed unused args for start/end frame (
f91ac03)
-
Remove area for ROI since approximations are probably a bad idea (
a54d88f) -
All options with dashes instead of underscores (
8dbc5ad)
-
Added tsne reduction (
d0f7647) -
Added square black pad resize for roi (
473aa34) -
Initial commit of new option to cluster roi only (
f08053c)
- Switch to multicore tSNE
(
9ec0ca9)
- Minor renaming for clarity
(
6bb998a)
- Added pass through of slicing overlap and postprocess_match_metric
(
f3b14bb)
- Adjust detection count per NMS output fix
(
9651c7a)
-
Allow for override of detect params with config.ini;correct save NMS output and detect single image (
4ab8780) -
More detail on cluster args, allow for override with config.ini, and set CUDA_VISIBLE_DEVICES in case not set (
40a96ef)
- Minor refactor of cuda device init
(
0f29e68)
- Sorted order needed to make start/end filtering work
(
9caea73)
- Erasing some prints in the code
(
622c119)
- Added support for specifying start/ending file names to support skipping image blocks
(
f344865)
- New features for clustering
(
b74d1b1)
- Deleting the os version
(
e63bb73)
- Adding .tiff images and vizualization tools with tqdm
(
d0b99e2)
- Added missing config.ini default
(
ae8b726)
- Bumped version
(
8ba97a8)
- Revised PYTHONPATH and switch to codfish semantic release
(
8eb6d42)
- Correct some bugs calling libaries
(
d69171b)
- Bumped version
(
404036b)
- release: 1.0.6 [skip ci]
(
ff2c3bd)
1.0.6 (2024-03-05)
- correct some bugs calling libaries (d69171b)
- Correct some bugs calling libaries
(
0668bda)
- Bumped version
(
b7a6740)
- release: 1.0.5 [skip ci]
(
1258290)
1.0.5 (2024-03-05)
- correct some bugs calling libaries (0668bda)
- Correct more imports
(
107a470)
- Bumped version
(
f1c17d3)
- release: 1.0.4 [skip ci]
(
5e76d32)
1.0.4 (2024-03-05)
- correct more imports (107a470)
- Correct import paths
(
6ca36e1)
- release: 1.0.3 [skip ci]
(
3dce93a)
1.0.3 (2024-03-05)
- correct import paths (6ca36e1)
- Added sahi example and constrain python to 3.9-3.11
(
240d630)
-
Add missing PYTHONPATH (
4371400) -
Minor typo fix in pytest.yml (
a108717) -
Fixed path to main and allow pytest to be run manually (
f586fd4)
- Bumped version
(
721b1b3)
- release: 1.0.2 [skip ci]
(
3d9488d)
1.0.2 (2024-02-17)
- Added spec removal ref
(
987ad8a)
- Added missing outer block for sahi pool
(
91f88ef)
- Bumped version
(
5508b33)
- release: 1.0.1 [skip ci]
(
f9f8639)
1.0.1 (2024-02-15)
- added missing outer block for sahi pool (91f88ef)
- Added release and pytest
(
1bb6f28)
- release: 1.0.0 [skip ci]
(
0c32f9b)
- initial commit (148f17f)
-
Correct links to image and reset version history (
2c98876) -
Added example images and cluster workflow diagram (
b228480)
- Initial commit
(
148f17f)