kwonjoon/Revisiting-Your-Memory

Revisiting Your Memory: Reconstruction of Affect-Contextualized Memory via EEG-guided Audiovisual Generation

ACM Multimedia 2025 CogMAEC Workshop (Oral)

Joonwoo Kwon*, Heehwan Wang*, Jinwoo Lee*, Sooyoung Kim*, Shinjae Yoo, Yuewei Lin, Jiook Cha

_{* equal contribution}

💻 Code on GitHub: github.com/kwxnjoon/Revisiting-Your-Memory  |  📄 Paper: arXiv 2412.05296

EEG-AffectiveMemory pairs EEG recordings during autobiographical memory recall with the memory's text description, a sketch, an associated musical piece and the participant's affect ratings. It supports the RevisitAffectiveMemory task, which reconstructs affect-contextualized audio-visual memories from EEG. This page holds the data. Code and the pretrained affect extractor are on GitHub.

Participants	Modalities	Size
10	EEG (raw + preprocessed), text, sketch, music link, affect	about 0.9 GB

Method

Files

Path	Content
sub-XX/raw/	Raw EEG of the baseline (base) and memory-recall (task) sessions (.edf, .easy, .nedf, .info)
sub-XX/pped/	Preprocessed EEG, three steps (EEGLAB .set / .fdt)
sub-XX/cebra_input/	Normalized input of the affect extractor (.csv)
sub-XX/sub-XX-decoded-matrix.csv	Decoded affect trajectory
sub-XX/sub-XX-sketch.PNG	Sketch of the recalled memory
sub-XX/sub-XX-text.txt	Memory description and the title and link of the associated musical piece (Korean)
text_descriptions.xlsx	All memory descriptions in one table

The musical pieces are not included because of copyright. Their titles and links are in sub-XX-text.txt.

Quick start

1. Download the data

hf download kwonjoon/Revisiting-Your-Memory --repo-type dataset --local-dir EEG_AffectiveMemory_dataset

2. Get the code

git clone https://github.com/kwxnjoon/Revisiting-Your-Memory.git

Place the downloaded folder at Revisiting-Your-Memory/EEG_AffectiveMemory_dataset/.

3. Run

• Affect extraction: CEBRA(Affect_Extractor)/ (pretrained model included in the GitHub repository).
• Image and video generation: image_video_decoding.ipynb.
• Music generation: put the melody at sub-XX/sub-XX-melody.wav, then run music_generation.ipynb.

License

Data under CC BY-NC 4.0 (non-commercial). The data come from human participants. Use them for research only and do not attempt to identify participants.

Citation

@inproceedings{kwon2025revisiting,
  title     = {Revisiting Your Memory: Reconstruction of Affect-Contextualized Memory via EEG-guided Audiovisual Generation},
  author    = {Kwon, Joonwoo and Wang, Heehwan and Lee, Jinwoo and Kim, Sooyoung and Yoo, Shinjae and Lin, Yuewei and Cha, Jiook},
  booktitle = {Proceedings of the 1st International Workshop on Cognition-oriented Multimodal Affective and Empathetic Computing},
  pages     = {1--10},
  year      = {2025}
}