Skip to content

Releases: iXab3r/YoloEase

0.1.65

22 Dec 22:40
Compare
Choose a tag to compare

Google Collab Pipeline

Fixed(it broke at some point) and improved Google Collab pipeline - it is now tested on 70GB archive of images and worked without any issues. The only real problem is how much time it takes to upload the data, for some reason basic upload via browser takes ages. What has greatly helped is uploading files to Google Drive and copying them via one of Notepad steps.

Also grouped all training-related steps into one section, so you can run them all
Run section

Notebook (you'll get that same link along with ZIP in the program)

Collab

Bugfixes/Improvements

  • Ultralytics Yolo no longer requires update. Yet to be checked, but I've added required variables accordingly to their docs
  • Fixed a problem with labels not being automatically pulled in
  • Fixed a bug which lead to UI not always being refreshed properly during Predictions stage
  • Fixed a bug with Run-once mechanism - it partially worked, you had to start-stop to give the program another run

0.1.63

17 Oct 09:29
Compare
Choose a tag to compare
  • Updated ONNX to 1.19.0
  • Added Yolo model name to final model name
  • Performance optimizations targeted at working with large datasets (20-30k+ images)

0.1.58

18 Aug 08:18
Compare
Choose a tag to compare
  • Added Yolo10 support
  • Added TrainVal percentage split
  • Few minor bugfixes

0.1.55

17 Aug 15:32
Compare
Choose a tag to compare

Augmentations

Augmentation performs transforms on your existing images to create new variations and increase the number of images in your dataset. This ultimately makes models more accurate across a broader range of use cases.

Augmentations tab
Augmentations dialog

0.1.54

16 Aug 22:58
9e61ef0
Compare
Choose a tag to compare

Enhancements to the Stage with YOLO Annotation

  • Now you can annotate not the entire remaining dataset, but only a part of it (%).
  • A new annotation mode has been added (now the default) – annotate only those files that have not yet been manually annotated.

Enhancements to Automatic Annotation

  • You can now specify a confidence threshold for each label individually.
  • You can set label priorities by dragging them in the list.

Enhancements to "Add Video"

  • You can now configure exactly which part of the video to extract.
  • You can specify the number of frames to extract.

Русская версия

Доработки этапа с разметкой через Yolo

  • теперь можно размечать не весь оставшийся датасет, а только его часть (%)
  • добавлен новый режим разметки(теперь по умолчанию) - размечать только те файлы, которые еще не были аннотированы вручную

Доработки автоматической аннотации

  • теперь можно указать порог(confidence) для каждой метки в отдельности
  • можно указывать приоритет меток перетаскиванием в списке

Доработки "Add Video"

  • теперь можно настраивать какой именно участок видео извлекать
  • можно указывать какое количество кадров извлекать

UI
Add Video

v0.1.36

26 Mar 22:31
Compare
Choose a tag to compare
  • fixed an issue with labels reordering
  • updated to NET8

Full Changelog: 0.1.19...0.1.36

YoloEase v0.1.19

18 Mar 23:08
882a147
Compare
Choose a tag to compare
Initial commit

YoloEase v0.1.17

18 Mar 19:17
882a147
Compare
Choose a tag to compare
Initial commit

YoloEase v0.1.16

18 Mar 17:36
882a147
Compare
Choose a tag to compare
Initial commit

YoloEase v0.1.11

16 Mar 12:46
882a147
Compare
Choose a tag to compare
Initial commit