fix: extend xlim in plot_ml_vs_dl to prevent clipping of value labels

feat: integrate wav2vec2 score into score fusion logic

fix: add wav2vec2 result handling and logging in analyze file processing

feat: add Wav2Vec2 detector service to analysis route

fix: adjust confidence calculation and streamline model download logic

fix: real ensemble inference, Youden threshold, DL unpickler

fix: update calibration and precision-recall plotting functions to load data from files

fix: update confusion matrix and score distribution functions to load pre-computed CV predictions

feat: add script to generate publication-ready figures for AURIS paper

fix: add ffmpeg decoding fallback for audio loading in analyze_vocals

fix: add ffmpeg decoding fallback for unsupported audio formats in feature extraction

fix: update evaluation to use optimal threshold and adjust BCEWithLogitsLoss for class imbalance

fix: adjust BCEWithLogitsLoss to handle class imbalance with pos_weight

fix: add missing roc_curve metric import for evaluation

feat: define families dictionary for candidate classifiers in _build_candidate_families function

feat: enhance candidate selection with class ratio and calibrated SVC for improved model performance

feat: implement optimal threshold calculation for binary predictions in train function

feat: add CalibratedClassifierCV and roc_curve imports for enhanced model evaluation

feat: add TorchSklearnWrapper for PyTorch model compatibility with sklearn and enhance final model training process

Add Gradio dependency to requirements.txt with version constraints

feat: train all models on full data and save their paths

Refactor code structure for improved readability and maintainability

feat: update duration limit for audio analysis to 1 minute for improved detection

feat: implement mixed precision training and gradient accumulation in training loop

feat: adjust confidence range in analysis result and optimize parallel processing for feature extraction, vocal analysis, and CLAP detection

feat: enhance score fusion confidence range and adjust CLAP weight for improved accuracy

feat: adjust confidence calculation in score fusion to improve agreement handling

feat: update model parameters for Random Forest, Gradient Boosting, XGBoost, and LightGBM to enhance performance

feat: adjust Random Forest parameters for improved model performance

feat: optimize Gradient Boosting and XGBoost parameters for improved model performance

feat: refactor training history figure generation to use GradientBoostingClassifier and improve metric tracking

feat: update model parameters and fix data leakage by removing duration_sec and sample_rate from features

feat: add deep learning classifier comparison script for AURIS with multiple architectures

feat: update excluded columns in feature statistics computation to prevent data leakage

feat: exclude additional metadata columns from feature loading to prevent data leakage

feat: update output directory for generated figures to docs/academic/figures

feat: add deep-learning figure generation script for enhanced model evaluation

feat: enhance calibration figure with Brier score and update prediction functions

feat: update ROC and Precision-Recall curve functions to include best model visualization

feat: implement cross-validated predictions caching and update confusion matrix function

feat: add additional imports for model evaluation and cross-validation

feat: add figure generation script for training results visualization

feat: enhance analysis result with estimated bitrate and XAI verdict integration

feat: add XAI inference for explainable predictions in analysis results

feat: add XAI explanation models for enhanced interpretability in analysis results

feat: integrate XAI service for enhanced explainability in analysis

fix: rename variable for clarity in compute_stats function

feat: add feature statistics computation from training CSV for enhanced analysis

feat: add XAI inference service for explainable predictions with detailed feature contributions

feat: implement resume functionality in batch feature extraction to skip processed samples