Loading...
Loading...
Contributors:
Dataset URI:
10.5281/zenodo.23227107Face-frame and spectrogram image sequences prepared from the Ryerson Audio-Visual Database of Emotional Speech and Song (RAVDESS) for the training and validation experiments in Yakaew et al., "Lightweight Audio–Visual Sentiment Analysis with Dynamic Mouth-State Modeling and Multimodal LLM Benchmarking" (under review). The dataset contains 569 audio-visual speech clips from 16 RAVDESS actors (01–07 and 11–19), grouped into three sentiment classes by valence: positive (happy; 128 clips), neutral (neutral and calm; 190 clips) and negative (sad and angry; 251 clips). Each clip has a sequence of face frames (12,140 JPEG images in total, 1280 × 720) and a sequence of spectrogram images (14,636 PNG images in total, 640 × 480) computed from the 48 kHz audio track. Audio segments with no sound produce blank spectrogram images, which are kept unchanged. No training/validation split is provided; a speaker-independent split is recommended. The files are organised by modality and class, and manifest.csv lists every clip with its RAVDESS emotion, intensity, actor and image counts. See README.md for full details, including the relation to Table 1 of the paper and known issues. This dataset is a derivative of RAVDESS and is released under CC BY-NC-SA 4.0. Please cite both RAVDESS (Livingstone and Russo, 2018, PLoS ONE 13(5): e0196391) and the paper above.
Keyword(s):
audio-visual; facial expression; RAVDESS; sentiment analysis; spectrogram.
MeSH Term(s):
Grants and Funding:
No Grants or Funding Data Available
Associated Publication(s):
No Associated Publication Data Available
Supplementary Data:
No Supplementary Data Available
Related Dataset(s):
No Related Dataset Data Available
The National Library of Medicine (NLM) provides information about biomedical datasets. Inclusion on an NLM Web page does not imply endorsement of, or agreement with, the contents by NLM or the National Institutes of Health. Learn more: NLM Web Policies (nih.gov)