- Bump pandas to 2.2.2 to fix broken pytest
(
32dc477)
- Removed unnecessary release step and minor changes to README.md to reflect correct package name
(
11f6558)
- Added missing upload to semantic release and poetry build command
(
e258cd3)
- Triggering release to test pypi
(
329bc28)
-
Clear names for dependency install and do release upload (
fa2b024) -
Replace pytest with poetry enabled (
9c45a21) -
Switching to poetry build (
419ca7d)
- Correct concurrent task array init
(
163eb25)
- 2x speed-up in crop
(
6c8f1fd)
- Correct option named not-verified to unverified
(
e97c2e2)
- Added flag to download all unverified data, e.g. --unverified python aidata download dataset --voc
--token --config /tmp/uav/config.yml --base-path /mnt/ML_SCRATCH/uav --min-score 0.5
--crop-roi --resize 224 --labels "all" --unverified
(
4dfa7da)
- Handle missing score and cluster attributes when combining versions
(
604c911)
- Download all versions by default and combine via NMS; this is potentially a breaking change for
downstream that relies on the Baseline naming convention. Remove Baseline from the root directory
name on datasets
(
7c8c1dd)
- Add database id and box coordinates to the localizations.csv output
(
9d948db)
- Handle missing secondary labels/scores during download
(
e1cec96)
- Added descriptions for just recipes
(
b187851)
- Added export of csv with media,cluster,label,score,label_s,score_s for downstream analysis
(
f2fe1fe)
- Larger batch size for ffmpeg roi crop
(
c3325b8)
- Correct related attribute query
(
cef33e8)
- Minor reorg and update docker info
(
c3915e4)
- Added support for minimum score in load boxes command and truncate exactly min boxes
(
3ed40b8)
- Remove any unused media attributes
(
0dee5d6)
-
Add missing arg for ffmpeg image crop that was munged during refactor (
7e9cb34) -
Rename plugins that load media; image_path should be media_path to support both image/video loads (
1d547e8)
- 2x speed-up for ROI crop
(
6d405c7)
-
Minor fix to assign mime first choice (
593f339) -
Minor fix to report correct name when video already loaded (
fbab8ee) -
cfe: Index video correctly for cfe isiis (
fae5709)
- Correct indexing for cfe media load
(
4a821af)
- Add moviepy==2.1.1 to support video metadata extraction
(
e2382f8)
- Minor update in load help for clarity
(
0691b85)
- Minor typo
(
0bef10a)
- Load CFE ISIIS video
(
15c236d)
- Reduce delay for redis load and more robust handling of different label payload
(
4abd444)
- Handle 2015-03-07T20:53:01.065Z media timestamp and different cases of label in redis load
(
e46a739)
- Separate crops by label and output stats.json to compatible format as voc-cropper
(
0c95b7e)
- Added support to download ROIs in optimized formats for training from either video or images with
python aidata download dataset --crop-roi --resize 224 etc.
(
a60c466)
- For coco/voc only download images and no video
(
2beb2e6)
- Remove any duplicate localizations on redis load
(
d6cd406)
- Correct handling of related attribute depth
(
190b3dc)
- Correct handling of related media attributes, e.g. --depth 1000 or --section 1000m during download
(
ff12787)
- Correct handling of related media attributes, e.g. --depth 1000 or --section 1000m during download
(
27d0ace)
- Exclude more than one label, e.g. --exclude Unknown --exclude Batray in load boxes
(
6f1fb51)
- Merge conflicts
(
ece4b53)
- cfe: Support loading individual cfe images
(
d0fea5d)
- Support any named exemplar
(
f097d7a)
- Added --max-saliency option to download
(
592439f)
- Handle missing float/int attributes
(
8c68ef9)
- Correctly load single voc/sdcat file
(
4e10f4b)
-
Minor update to command help to include voc (
a61d50d) -
Minor update to command help to include voc (
45177b8)
- Added support for collapsing to a single class during download with --single-class
(
72b2aa1)
- Handle missing video frame rate/codec, more performant queue based loads and handle different
timecode payloads
(
876036a)
- Added update to README.md
(
7330522)
- Added support to load from voc
(
3b653ec)
- Minor typo fis
(
7a75bee)
- Support exclusion of single label, e.g. Unknown with load boxes
(
3be99be)
- Support redis port/host/password
(
3a25060)
- uav: Correct media date
(
76bf902)
- Add support for --min-score in download dataset
(
915aa1f)
- Albumentations dependency break
(
b483324)
- Add support for --min-saliency in download dataset
(
014f5af)
- Slimmer memory footprint for downloads. working for 360k download and export for CFE ROIs
(
4f035ac)
- Added support for --section in download
(
a89e695)
-
Add missing images (
4a2171e) -
Correct link to docs (
2c7a544) -
Just some refactoring of setup docs (
6531f40)
Adds the option --resize to the transform to transform with a resize. Useful for training models with scales of the same image. For tiny boxes that may result in downsizing, the option --min-dim supports removal at them; this defaults to 10 pixels in any dimension
-
feat: added test and config for loading planktivore
-
docs: more detail on test/dev setup and added in redis to test
-
test: correct tests conf that aligns to doc
-
docs: correct link to database
-
chore: added in more just recipes for testing
- Add support to different models
(
13fb78e)
- Bump to better indexing
(
5405abe)
- Replace underscore with colon in redis hash store of exemplars
(
d7a07bc)
- Added redis document id fetch
(
3ab3b72)
- Replace index id with database id for exemplars
(
64f6dde)
- Disable flagging exexmplars
(
fa74c55)
-
Updated docs and and config with better test setup (
f45e94f) -
Updated docs and and config with better test setup (
7210a84)
- Added support for depth parsing
(
9449cde)
- Handle empty query operators
(
b957f7a)
- Trigger release to update init.py
(
d278ce3)
- Minor correction to reflect correct path to docs
(
13c24b5)
-
Correct handling of label map (
7827810) -
Handle conversion errors outside of normalized 0-1 coordinates (
e410cd4)
- Added missing albumentations library
(
623df2f)
-
Added detail on .env file (
66b83a0) -
Correct link to internal docs (
c0e4f87) -
Moved CHANGELOG (
f3df55c) -
Removed docs which are now in another repo and added commands for ref in README.md (
bdd122a) -
Resolve conflict (
783fa25) -
Updated with correct link to docs (
1577fa3)
-
Addded voc_to_yolo (
cb4e798) -
Added transform for voc only (
ade3713) -
Added transform for voc only (
9147c28)
- Added label counts to /labels/{project_name}
(
d6e5cce)
- Added verified --verified to download
(
bb2b274)
- Added database reset and cleane up docs
(
572bdd5)
- Add load to exemplar bool which is helpful for visualization
(
5d1feb2)
- Correct default for max-images to load all
(
9f284ab)
- Correct parsing of class names and cuda device enable
(
58b0dea)
- Correct handling of exemplars
(
accbb9e)
-
Correct docker build path (
4406a96) -
Exclude py12 which is problematic with transformers library (
2116ad3) -
Remove legacy "ENV key value" in Dockerfile.cuda (
84fa73c)
- Correct handling of version arg
(
8eb99b5)
- Added password pass through to redis
(
900bbef)
- Move model to gpu and some minor refactoring for clarity
(
a9d2f59)
- Minor fix in exemplar args and more logging
(
f34e643)
- Handled empty labels, bad media, and more reporting of progress
(
e31b691)
- Bugs from typecheck
(
222d641)
- Removed unused imports and bump torch to python3.11 compatible and
(
c9afcbb)
- Remove whitespace and move exemplar to load
(
a7d8607)
- Added sdcat exemplar load
(
46ea57e)
- Improved doc on save
(
7c37c42)
- Pass through the id to the voc output
(
e28a211)
- Handle fewer records correctly
(
a3528de)
-
Correct load_bulk_boxes args and attribute mapping for redis load (
09241be) -
Revert to original pass through of datetime object (
5af3119)
-
Handle variable case attribues (
2538007) -
Remove audio and reduce frame rate to 24 for .git (
604a463) -
Remove palette gen for speed-up of video gif creation (
5ce9a12)
- Speed-up 64x speed
(
f0cd15c)
- Correct key for datetime attribute format
(
f1bcad7)
- Added in load/download docs and adding in CHANGELOG.md
(
f3ac92c)
- Correct import path
(
0206f3d)
- Add missing files
(
b95710a)
-
Minor fix to checkout and more explicit naming of path (
7be3aa3) -
Minor update to add token to examples (
7c499a5)
- Some refactoring but mostly addition of support for versioning
(
cfe9ba1)
- Adjustments to match Fernandas generated test images
(
1ef27b7)
- Added support to pass in max-images which is useful with --dry-run
(
fe8d2de)
- Remove return to load all
(
460b449)
- Skip over images with no metadata and minor logging fix
(
c926ca8)
-
Correct path for loading SONY images (
42048fb) -
Return dataframe and process in sorted order for convenience (
52025fb)
- Added missing dependency
(
3f1a267)
- Added hint for dry-run and hostname
(
f314d4e)
- Minor type
(
f70d7bd)
-
Support both png and jpg SONY images (
75adcca) -
Support both png and jpg uppercase SONY images (
343651b)
- Added support for extracing sony mdata for UAV
(
6cd2faa)
- Correct path to data in test database
(
fa9f7dc)
- Correct formatting for the cluster string
(
41678b3)
- Fix bug that downloads everything
(
685c72a)
- Initial commit
(
9ce6cc4)