Acknowledgement We would like to express our sincere gratitude to the creators of the EGDB dataset for providing such a valuable resource for our research. The EGDB dataset, as detailed in the paper “Towards Automatic Transcription of Polyphonic Electric Guitar Music: A New Dataset and a Multi-Loss Transformer Model” by Chen, Yu-Hua, et al., provides a comprehensive collection of guitar tracks that enabled us to evaluate our models. Their efforts in curating and making this dataset available have significantly contributed to advancements in the field of guitar effect removal. Reference: Chen, Yu-Hua, et al. “Towards Automatic Transcription of Polyphonic Electric Guitar Music: A New Dataset and a Multi-Loss Transformer Model.” ICASSP 2022-2022 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). IEEE, 2022. We appreciate the dedication and hard work put into creating and maintaining the EGDB dataset, and we look forward to its continued use in future research endeavors.

This dataset is used for additional experiments for our work: Distortion Recovery: A Two-Stage Method for Guitar Effect Removal, published at DAFx 2024. We randomly rendered dry (direct input) guitar tracks from the EGDB dataset using BIAS FX2 and fine-tuned both our model and Demucs V3 on this rendered dataset. The EGDB dataset consists of 240 tracks (~1.5 hours) and was randomly split by tracks into training, validation, and test sets, following an 8/1/1 split. Our model was then compared with Demucs V3. During both the training and testing phases, each track was randomly segmented into 4-second clips. For reproducibility, the segmented clips for testing, including wet/dry signal, results of our model and Demucs V3, are available here. The test code for evaluating the following metrics is available at this GitHub repository.