Deep Learning / Affective Computing — Tokyo / Chiba
Teaching machines how a body feels
Can a camera tell how stressed you are? That question is my research in AI and deep learning; this blog is its trail — a worklog, a shelf of writing, proof that I am still working.
LIVE — REMOTE PPG, RECONSTRUCTED≈ 62 bpm
Research one line, three papers
UBFC-PHYS · TSST-INSPIREDPublishedEDA-augmented CCT-LSTM for stress-task classificationThe first public stop on this line — contact signals join the video.KICS · short paperAbsorbedA reliability audit of remote-PPG under social stressWhy contact signals stay: rPPG degrades exactly when stress happens.audit → journal, §motivationUnder reviewThe journal version — reproducible, fully openNew architecture, strict protocol, full ablations.IEEE J-BHI · target
Protocol7-fold × 5 seedssubject-level splits, never sample-level
paired statistics, corrected
paired statistics, corrected
multimodal video · reproducibility · temporal modeling · next: VLM / MLLM for affect
Writing essays · poems · research notes
all writing →« J’aime les nuages… les nuages qui passent… là-bas… là-bas… les merveilleux nuages ! »
I love the clouds… the clouds that pass… up there… up there… the marvelous clouds!
— Baudelaire, L’Étranger FR