Overview
Version **v0.6.0** brings an array of enhancements, bug fixes, and new features, improving the functionality, usability, and robustness of the **Cucaracha** library. This release focuses on advancing machine learning tasks such as image classification and segmentation, data augmentation, and model evaluation, while refining the codebase and documentation.
---
🚀 New Features
Image Classification Enhancements
- **Dataset Augmentation:**
- Added a script to generate augmented datasets.
Commit: `a990cc1`
- Improved data augmentation techniques.
Commit: `54ec923`
- **Visualization Support:**
- Added functionality to visualize training samples before proceeding with training.
Commit: `bcad26a`
- **Custom TensorBoard Callback:**
- Log training images during model training.
Commit: `f069df4`
- **Architecture Listing:**
- Added `--arch_list` option to display available classification architectures.
Commit: `ee33725`
- **Evaluation Script:**
- Introduced a detailed reporting and visualization script for evaluating pretrained models.
Commit: `07c734d`
Image Segmentation Enhancements
- Refactored `ImageSegmentationTrainer` for better model initialization and metrics handling.
Commit: `35d297a`
Data Handling
- Added support for loading class labels from a JSON file in the model evaluation script.
Commit: `242e9da`
- Integrated the `image_auto_fit` function to resize images for Keras model input.
Commit: `e03a3c2`
Dependency Updates
- Added `scikit-learn` and `seaborn` for data visualization and analytics.
Commit: `df8f783`
- Upgraded `keras` to version 3.8.0 and included `scipy` for scientific computing.
Commits: `294b476`, `d7adf65`
---
🛠️ Enhancements
- Updated parameter validation for image classification tasks.
Commit: `588b401`
- Introduced functions to plot confusion matrices and handle organized datasets efficiently.
Commits: `cf46b69`, `fca10ff`
- Improved configuration files with validation for allowed image extensions.
Commit: `ce666fb`
---
🐛 Bug Fixes
- Fixed import paths to align modules and tasks.
Commits: `b63d944`, `a039329`
- Fixed `__str__` method implementations for ML models.
Commit: `ca35d5b`
- Resolved issues with missing initialization files in the `tasks` module.
Commit: `57f8f4a`
- Removed unused image file links from classification datasets.
Commit: `2c59293`
---
📝 Documentation Updates
- Improved docstrings for better clarity and formatting in key functions.
Commit: `2075c1e`
- Added detailed documentation for model architect and identification functions.
Commit: `0129375`
---
🔧 Codebase Improvements
- Refactored import statements for consistency and readability.
Commits: `8f261c8`, `37be488`
- Improved code style with pragma comments and clean-up tasks.
Commits: `acdefe4`, `7d33184`
---
Upgrade Instructions
1. **Update the library:**
bash
pip install cucaracha --upgrade
2. **Explore new features:**
Check out the updated [documentation](https://cucaracha.readthedocs.io/en/v0.6.0/) for guidance on using the new features.
Enjoy the new version and happy coding! 🚀