v2.3.0

What's Changed

Known issues

• The audio file selection dialog in the single file tab doesn't allow the selection of files on macOS. But using drag-drop still works.
• Exported Raven models, currently don't run in Raven Pro, likely due to changes in our Tensorflow version and the way the model is created. Please use a previous version if you want to continue training Raven models for now.

Fixes

• Fixed bug that saves faulty custom classifier if 'both' was selected by @Josef-Haupt in #783
• Fixed files not being saved to disk by @Josef-Haupt in #796
• Fix focal loss in autotune and apply sigmoid for test data evaluation by @max-mauermann in #792
• Monitor validation loss for early stopping by @max-mauermann in #793

Changes

• Move embedding-search to tab bar by @Josef-Haupt in #784
• Fail on empty line in species list + trim by @Josef-Haupt in #790

Features

• Added Perch v2 as an alternative by @Josef-Haupt in #788

Full Changelog: v2.2.0...v2.3.0

v2.2.0

What's Changed

• Order of additional columns in the csv is now always consistent by @max-mauermann in #756
• Fixed focal loss not beeing optimized by autotune by @Josef-Haupt in #759
• Improve embedding extraction speed by @max-mauermann in #736
• Fix labels not being read when using a custom classifier by @Josef-Haupt in #767
• Only output skipped files for batch analysis by @Josef-Haupt in #770
• Added three segment collection modes by @max-mauermann in #731
• Support for multiprocessing for embedding extraction in the embeddings-extract tab by @Josef-Haupt in #774
• Fixed bug that displayed the collected files in the batch analysis tab incorrectly by @Josef-Haupt in #779

Full Changelog: v2.1.1...v2.2.0

v2.1.1

What's Changed

• Fix the GUI version update checker by @emmanuel-ferdman in #716
• Allow to run custom classifier files without the labelsfile by @Josef-Haupt in #719
• Escaped "<" and ">" in GUI by @Josef-Haupt in #720
• Update Serbian translation by @zemzema in #723 and #714
• Fixed timestamps not respecting overlap when using different audio speed by @Josef-Haupt in #722
• Streamline chunk generation for embeddings and analyze by @Josef-Haupt in #727
• Move to two digit precision in results to improve accuracy when using overlap+audiospeed by @Josef-Haupt in #730

New Contributors

• @emmanuel-ferdman made their first contribution in #716

Full Changelog: v2.1.0...V2.1.1

v2.1.0

What's Changed

• Added deepl translations for the dog label by @max-mauermann in #668
• Fixed audio speed not iterating whole file by @Josef-Haupt in #678
• Added up to date eBird taxonomy codes + fixed bug in train CLI by @Josef-Haupt in #683
• Fixed bug that could in edge cases lead to wrong embeddings beeing extracted by @Josef-Haupt in #691
• Fixed bug preventing download of table in single-file tab by @max-mauermann in #696
• Added the option for additional metadata to the csv output format by @Josef-Haupt in #692
• Russian translation updates by @Josef-Haupt in #700
• Added support link to footer by @Josef-Haupt in #706
• Plots in evaluation tab get cached now by @Josef-Haupt in #705
• Added updates for labels on Serbian language by @zemzema in #688
• Brought back the option to output files for embeddings by @max-mauermann in #679

New Contributors

• @zemzema made their first contribution in #674

Full Changelog: v2.0.0...v2.1.0

v2.0.0

Breaking

• Remove r table output type by @Josef-Haupt in #549
• Rework CLI args (see documentation) by @Josef-Haupt in #564
• Change how sensitivity works by @kahst in #578 and #579

What's Changed

• Save batch analysis args by @Josef-Haupt in #548
• Allow easier usage of old BirdNET models by @Josef-Haupt in #550
• Add audio speed modification for ultra- and infrasonic recordings by @kahst in #557
• Add top_n argument, to only save the top N predictions for each segment by @Josef-Haupt in #577
• Added map to species selection tab by @Josef-Haupt in #583
• Added review tab keyboard shortcuts by @Josef-Haupt in #582 and #591
• Add arg to merge consecutive detections by @kahst in #593
• Embeddings with hoplite database by @max-mauermann in #562
• Startup performance by @Josef-Haupt in #599
• Single file tab now gives error messages for invalid audio files by @max-mauermann in #603
• Indonesian language support for GUI by @Josef-Haupt in #605
• Updated Finnish labels by @jarnoh in #629
• Added option to switch to dark theme by @Josef-Haupt in #621
• Finnish and Swedish translations for GUI by @tphakala in #631
• Better error logging in GUI + CLI by @Josef-Haupt in #639
• Added hidden file exclusion for MacOS in review and segments by @yurish in #635
• Fixed bug that prevented the error_log to be written to userdir by @Josef-Haupt in #641
• Add known network/cloud shortcuts to allowed paths on windows by @Josef-Haupt in #642
• Performance app by Rene Heinrich and @Josef-Haupt in #645
• The locale selection list now respects locales like en_uk and pt_br by @max-mauermann in #643
• Allow result download + segment download in single file tab by @Josef-Haupt in #647
• Conditional parameters in tuner by @max-mauermann in #421
• Allow to use test data for training by @kahst in #652
• Improve custom classifier training by @kahst in #654

New Contributors

• @cbowns made their first contribution in #540
• @jarnoh made their first contribution in #629
• @yurish made their first contribution in #627
• @tphakala made their first contribution in #631

Full Changelog: v1.5.1...v2.0.0

v2.0.0-rc

Breaking

• Remove r table output type by @Josef-Haupt in #549
• Rework CLI args (see documentation) by @Josef-Haupt in #564
• Change how sensitivity works by @kahst in #578 and #579

What's Changed

• Batch analysis args are now saved in the output directory by @Josef-Haupt in #548
• Allow easier usage of old BirdNET models by @Josef-Haupt in #550
• Add audio speed modification for ultra- and infrasonic recordings by @kahst in #557 and #561
• Add top_n argument, to only save the top N predictions for each segment by @Josef-Haupt in #577
• Added map to species selection tab by @Josef-Haupt in #583
• Added review tab keyboard shortcuts by @Josef-Haupt in #582 and #591
• Add arg to merge consecutive detections by @kahst in #593
• Embeddings with hoplite by @max-mauermann in #562
• Better startup performance by @Josef-Haupt in #599
• Single file tab now gives error messages for invalid audio files by @max-mauermann in #603

Known issues

• Progressbar not displayed correctly

Misc

• Add new documentation by @Josef-Haupt in #564 at https://kahst.github.io/BirdNET-Analyzer/

New Contributors

• @cbowns made their first contribution in #540

Full Changelog: v1.5.1...v2.0.0-rc

v1.5.1

What's Changed

• Download buttons for regression and spectrogram plots in review tab by @Josef-Haupt in #529
• correct audio segment gets played in single file tab result by @Josef-Haupt in #530
• updated server.py and client.py to work with current analyze.py by @Josef-Haupt in #532

New Contributors

• @Olindn made their first contribution in #517
• @kenshih made their first contribution in #526

Full Changelog: v1.5.0...v1.5.1

v1.5.0

What's Changed

• fixed the update notification by @Josef-Haupt in #505
• added optional spectrogram to single file analysis tab by @Josef-Haupt in #506
• more readable font colors by @Josef-Haupt in #507
• added ascii logo to analyzer by @Josef-Haupt in #500
• updated russian translation by Александр Цветков by @max-mauermann in #508
• Single file tab plays selected segments in result table by @Josef-Haupt in #509
• portuguese labels are now actually portuguese and not english (Thanks @Mattk70) by @Josef-Haupt in #515
• better GUI exception for empty training classes by @Josef-Haupt in #513
• allow single class review in review tab by @Josef-Haupt in #520

New Contributors

• @kdarras made their first contribution in #503
• @tvoirand made their first contribution in #514

Full Changelog: 1.4.0...v1.5.0

v1.4.0

What's Changed

• Add portuguese-BR translations to the GUI by @LSMSugai in #439
• Russian GUI translation by Александр Цветков by @max-mauermann in #448 (Thanks to Alexander-R1BAF)
• Fixed append-training in installed version of the GUI by @max-mauermann in #456
• writing the states and settings jsons is now more robust by @max-mauermann in #457
• changed single analysis result header to hh:mm:ss format by @max-mauermann in #460
• in GUI batch-analysis the order of the result files is now aligned with the order of the input files by @max-mauermann in #473
• the GUI now resets the loaded custom classifier before starting the analysis by @max-mauermann in #473
• model params after training now include the model version + an additional csv containing the number of samples for each class-combination by @max-mauermann in #473
• native windows directory selection dialog by @nishantnnb in #473
• updated chinese labels by @RockerLee in #473
• Gui refactor by @Josef-Haupt in #473
• added labels for bg, ca el, he, hr, in, is, lt, ml, sr by @Josef-Haupt in #480
• single file analysis output can display utf-8 characters by @Josef-Haupt in #482
• Updated to Gradio 5 by @Josef-Haupt in #489
• added some validation to the gui and fixed folder selection issues for training cache by @max-mauermann in #491

New Contributors

• @LSMSugai made their first contribution in #439
• @tommy4st made their first contribution in #492

Full Changelog: v1.3.1...1.4.0

v1.3.1

What has changed

• Fixed the bug in segments that prevented using audio and result files from different folders
• Fixed the bug that showed the splash-screen during analysis in the windows version
• Fixed the bug that audio could not be played in the mac version