[1.11.0-alpha] - 2022-12-20

🚀 Added

• VOC XML import (https://github.com/SkalskiP/make-sense/pull/280). Now, we can both import and export annotations in VOC XML format.
• Support for Roboflow inference server (https://github.com/SkalskiP/make-sense/pull/298). Now, you can use Roboflow API to annotate images in MS.

https://user-images.githubusercontent.com/26109316/208692860-77fb999e-dd3b-4e23-9fca-c3cfa4751dd5.mp4

🔧 Fixed

• Fix support for YOLOv5 P6 models (https://github.com/SkalskiP/make-sense/issues/281). We have resolved problems with loading YOLOv5 M6 models into MS. These were related to the required higher inference resolution.
• Fix confusion around UI for uploading labels (https://github.com/SkalskiP/make-sense/issues/282). UI fixes to make it easier for users to use YOLOv5 models in MS.
• Fix non unique label names handling (https://github.com/SkalskiP/make-sense/issues/286). So far, MS has not allowed the user to have multiple labels with the same name, but has not explained in any way why it refuses to load the project. Now an appropriate message with instructions is displayed.
• Fix docker build fails (https://github.com/SkalskiP/make-sense/issues/289). After JS dependencies changes, it was necessary to bump up the required version of the base node image.

🏆 Contributors

@hartmannr76, @PawelPeczek, @SkalskiP

🚧 Work in progress

• Our work on stats view is almost done and you can expect it in the next release.
• We are working on integration with MS Inference Server developed courtesy of @PawelPeczek.

[1.10.0-alpha] - 2022-09-30

🚀 Added

• Support for YOLOv5 models. (https://github.com/SkalskiP/make-sense/pull/276) You can now use pre-trained or custom models to speed up the annotation process. Take a look at the short video highlighting the new feature.

https://user-images.githubusercontent.com/26109316/193255987-2d01c549-48c3-41ae-87e9-e1b378968966.mov

🔧 Fixed

• Fix UI bugs that allowed to interact with hidden labels (https://github.com/SkalskiP/make-sense/pull/272)

🏆 Contributors

Special thanks are due to @IuliuNovac for his original PR showcasing integration with YOLOv5. We most definitely would not have that killer feature in makesense.ai without your spark. 💥 Thank you! 🙏

🚧 Work in progress

One of the definite highlights of the next release will be the introduction of the long-awaited stats view, which would present statistics about annotated objects and images. The work on that feature is already quite advanced. (https://github.com/SkalskiP/make-sense/pull/268)

What's Changed

• Add twitch link to Make Sense landing page by @SkalskiP in https://github.com/SkalskiP/make-sense/pull/270
• fix/it_is_possible_to_interact_with_hidden_labels by @SkalskiP in https://github.com/SkalskiP/make-sense/pull/272
• feature/156_import_custom_model_yolov5 by @SkalskiP in https://github.com/SkalskiP/make-sense/pull/276
• 1.10.0-alpha relese by @SkalskiP in https://github.com/SkalskiP/make-sense/pull/277
• 1.10.0-alpha relese by @SkalskiP in https://github.com/SkalskiP/make-sense/pull/278

Full Changelog: https://github.com/SkalskiP/make-sense/compare/1.9.0-alpha...1.10.0-alpha

[1.9.0-alpha] - 2022-07-26

🚀 Added

• Turning annotation visibility on and off (https://github.com/skalskip/make-sense/issues/255)
• Introduction and configuration of the linter (https://github.com/SkalskiP/make-sense/pull/262)
• Adding headers to CSV files created as a result of exporting box annotations (https://github.com/skalskip/make-sense/issues/252)

https://user-images.githubusercontent.com/26109316/181109414-a96382bf-75b1-48a8-962c-528315136934.MP4

🌱 Changed

• Migrate to the latest versions of React and MaterialUI (https://github.com/SkalskiP/make-sense/pull/262)
• Migrate from Webpack to Vite (https://github.com/SkalskiP/make-sense/pull/262)

🔧 Fixed

• Fix docker installation problems (https://github.com/SkalskiP/make-sense/issues/210)
• Fix PASCAL XML export to work with TFLite Model Maker (https://github.com/SkalskiP/make-sense/pull/214)

🏆 Contributors

Special thanks are due to @rasyidf for his exceptional activity, both in the form of PRs, as well as great ideas for feature development. Than you! 🙏

@rasyidf, @IuliuNovac, @Parthiban

🚧 Work in progress

One of the definite highlights of the next release will be the introduction of the ability to load any custom YOLOv5 model into MS, and use it when labeling images. Thanks to @IuliuNovac's tremendous work, we already have an advanced and working proof of concept for this feature (https://github.com/SkalskiP/make-sense/pull/220).

[1.8.0-alpha] - 2021-09-06

🚀 Added

• GitHun action with greeting message triggered when opening each new issue (#182)
• Automated tests and integration with CodeCov (#184)
• Tooltips for editor action bar buttons (#196)
• Selection of label name via keyboard shortcuts (#134)
• First edition of make-sense documentation (#185) - https://skalskip.github.io/make-sense/
• Per class annotation coloration (#190) ⭐
• Centralised error notification system (#193) ⭐

🌱 Changed

• Bump node version and npm dependencies (#189 and #191)
• Sort images by name during image loading (#195)

🔧 Fixed

• Importing images and labels with filenames containing multiple dots (#160)
• Invalid YOLO annotations exported for bboxes on the edge of an image (#188)
• Starting a line outside of the image causes an error that knocks out the entire editor (#194)

🏆 Contributors

@jaaywags, @rasyidf, @h3ct0r, @Minipada and @SkalskiP

• Faster labels definition - use the Enter key to confirm currently edited label and open a new text field
• Updated external dependencies
• Fixed bugs around mouse crosshair
• Style changes - introducing a dropdown in the application top bar
• Importing annotations in COCO and YOLO formats
• Exporting annotations in COCO format
• Simplifying popups management
• Adding new unit tests

• Image classification labeling https://github.com/SkalskiP/make-sense/issues/81
• Using lines to label images https://github.com/SkalskiP/make-sense/issues/84
• Extended cross-hair https://github.com/SkalskiP/make-sense/issues/12

• Improved integration with the SSD model - now, apart from bboxes, we have added a suggestion of the class to which the object belongs
• Integration with the PoseNet model - can be used to estimate the pose of a person in an image or video by estimating where key body joints are
• #15 Add and rename Labels after creating
• Many smaller bugs have been eliminated

Added integration with SSD model pretrained on the COCO dataset, which will do some of the work for you in drawing bboxes on photos and - in future versions of the application - will also suggest a label.

• Zoom in and out of photos and the ability to manimpulate a close-up photo
• Support for new keyboard shortcuts and patches in the existing shortcuts system
• Fixing of application bugs, among others in the scope of window scaling

Due to the large amount of bugs that we have managed to fix recently, we decided to make a small release. It was important because in the near future we will spend a lot of time on developing new functionalities, and at the same time we wanted to keep the application in the best possible form.

Among other things, we managed to fix:

• #36 If there is only points annotation, the exported csv contains nothing.
• #37 No file name in the csv file if more than 1 images in the project
• #40 Scrolling label
• #43 Visual bug