v2.0.0b1

BatDetect2 2.0.0b1

This is the first beta release of the new batdetect2 2.0 workflow.
2.0.0b1 is a pre-release. We would really appreciate feedback while the new CLI, Python API, training, evaluation, and output workflows settle.

Highlights

• New BatDetect2API workflow for inference, training, evaluation, and fine-tuning
• New process CLI workflow for batch inference
• Built-in default checkpoint handling and easier checkpoint loading
• Support for loading checkpoints from Hugging Face
• Fine-tuning from checkpoints
• Expanded output format support, including raw, batdetect2, soundevent, and parquet-style workflows
• Restored legacy-style batdetect2 process outputs where needed
• Configurable class labels and richer target/config handling
• Refreshed docs, tutorials, references, and migration guide
• Refreshed default bundled checkpoint/model

Docs

• Start in docs/
• Migration help: docs/source/legacy/migration-guide.md

Feedback

If you run into issues or have feedback on the new workflows, please open a GitHub issue.

v1.3.1

What's Changed

• Http documentation by @kaviecos in #52
• Remove version cap for torch and torchaudio, add Python 3.13/3.14 support by @mbsantiago in #61

Full Changelog: v1.3.0...v1.3.1

v1.3.0

What's Changed

• Http support by @kaviecos in #44

New Contributors

• @kaviecos made their first contribution in #44

Full Changelog: v1.2.0...v1.3.0

v1.2.0

What's Changed

• Added the chunk_size param to the detect command by @mbsantiago in #45

Full Changelog: v1.1.1...v1.2.0

v1.1.1

What's Changed

• fix: Address PyTorch Model Loading Deprecation Warning (GH-30) by @mbsantiago in #37
• test: Add failing audio files from GH-29 to contrib test suite by @mbsantiago in #38
• fix: Handle Empty Audio Files Gracefully (GH-20) by @mbsantiago in #39

Full Changelog: v1.1.0...v1.1.1

v1.1.0

What's Changed

• Fix numpy exception handling by @ccarrizosa in #32
• Feat/migrate to uv by @mbsantiago in #33
• Feat/migrate to numpy 2 by @mbsantiago in #34
• feat: Drop Python 3.8 Support, Add Python 3.12 Support by @mbsantiago in #35
• fix: Resolve detect Command Failure with Specific Audio Files (GH-31) by @mbsantiago in #36

New Contributors

• @ccarrizosa made their first contribution in #32

Full Changelog: v1.0.8...v1.1.0

v1.0.8

Release Notes

This patch release focuses on maintenance, addressing compatibility. Key updates include:

1. Package Dependency Compatibility: The minimum version requirements for package dependencies have been adjusted. batdetect2 is now compatible with PyTorch 2.

2. Google Colab Compatibility: batdetect2 can now run in Google Colab without the necessity of installing custom versions of machine learning packages. As of 30 Jan 2024.

3. Extended Python 3.11 Support: We've expanded our support to include Python 3.11.

v1.0.7

Release Notes

This patch addresses Issue #22, where batdetect2 was inadvertently outputting detections above the Nyquist limit. We appreciate @kdarras for bringing this to our attention.

To rectify the issue, a straightforward cleaning step has been incorporated to eliminate detections surpassing the Nyquist limit.

Special thanks again to @kdarras for graciously providing one of his recordings for testing the modification. The recording has been placed in the tests/data folder (file). Additionally, a new test/test_detections.py file has been created to implement tests ensuring that batdetect's outputted detections are free from any significant errors.

v1.0.6

Release v1.0.6

Spectrogram Feature Computation Issue Fix and Code Refactoring

In this release, we've addressed a critical issue related to the computation of spectrogram features, marked by the generation of erroneous csv output containing -1 values. This problem stemmed from an unanticipated shape discrepancy within the spectrogram, leading to improper feature extraction.

What's New

• Bug Fix: The spectrogram feature computation issue has been resolved. The csv output, which was previously populated with -1 values, now accurately reflects the calculated spectrogram features.
• Code Refactoring: To improve maintainability and extensibility, the feature extraction functions have undergone a thorough refactoring. This modularization enhances the organization of the codebase and facilitates more efficient testing and future enhancements.

• Unit Testing: With the introduction of modularization, each feature extraction function is now independently unit tested. This ensures the reliability and correctness of these functions in isolation.
• Integration Testing: An integration test has been added to verify the correct functioning of the CLI command with the "--spec_features" flag. This integration test validates the end-to-end behavior of the feature extraction process.

v1.0.5

Bug Fixes:

1. Issue #10 was fixed by making sure empty files or files without predictions are handled correctly. Added a new test case for this scenario. Modified some of the functions that post-process the model's prediction to handle the case of no predictions adequately.