Skip to content

Odin v1.1.0

Latest
Latest

Choose a tag to compare

View all tags
@matjsz matjsz released this 29 Dec 00:30
· 14 commits to main since this release
Latest
634c7c8

Changelog

✨ New | Model automated training command interface for training pipelines

This change includes a new command interface (model) that contains three main functionalities:

  • train: Automatically trains a new model, it can be categorized as a naive or a wise model, if the model is trained on a subset, then it's a naive model, otherwise, it's a wise one.
  • test: In this update, the test command will be able to load a specific version of a chronicle and test on the validantion split of the chronicle's dataset, the plan is to have this as part of the pipeline and automatically run real tests, comparing detections and classifications to the ground-truth expected values, running as a "Cypress for computer vision AI"
  • publish: Serves a version controller, publishes the model to the folder where final versions of models in the specified project are stored. The models are versioned by project, date of training and chonicle.

This feature is available for both Object Detection and Classification projects.

Fix: 🛠 | Dataset command interface UX improvements

In this update, several UX improvements and fixes has been made to the dataset interface:

  • Fixed the problem with custom classification datasets not generating folders for the custom classes
  • Fixed the requirements of the interface, now all arguments were converted to options, if an option is not included in the command, Odin will ask for its value if it's truly necessary, making it more dynamic
  • Error handling and folder creation checks were applied so Odin can figure out what to do in bizarre situations, there are still some bugs that may happen, but those will be fixed as the project evolves

Fix: 🛠 | Basic commands bug fixes

Some commands received quality of life upgrades so minor maintenance will be needed.

Odin's Launch! 🎉

The tool is finally being released on PyPI! After these last updates, the core functionalities are done, the rest is extra and will be added in the next updates, but the tool will be generally available at Python's official repository for community-made packages.

The official ID of the package will be at the README of the repository.

Thanks for reading! :)

Assets 4
Loading