Task
General
L
mandipgoswami/LibriRIR-100
General · mandipgoswami· 98
cc-by-nc-4.0
21 GB
LibriRIR-100 is a large-scale paired clean↔reverberant speech training corpus containing exactly 100 hours of speech. Each utterance is paired with a room impulse response from RIR-Mega (mandipgoswami/rirmega), stratified across four RT60 reverberation conditions. Designed as a drop-in training resource for robust ASR, speech enhancement, and dereverberation models.
# download instantly
mlforge datasets pull mandipgoswami/LibriRIR-100
mlforge datasets pull mandipgoswami/LibriRIR-100
Dataset details
Language
en
License
cc-by-nc-4.0
Size
21 GB
Rows / images
28.4K
Creator
mandipgoswami
Downloads
98
Source
huggingface_datasets
Updated
2026-03-04
About mandipgoswami/LibriRIR-100
LibriRIR-100 is a large-scale paired clean↔reverberant speech training corpus containing exactly 100 hours of speech. Each utterance is paired with a room impulse response from RIR-Mega (mandipgoswami/rirmega), stratified across four RT60 reverberation conditions. Designed as a drop-in training resource for robust ASR, speech enhancement, and dereverberation models.