• keepthepace
    link
    fedilink
    Français
    arrow-up
    2
    ·
    4 months ago

    Small input/output windows. The ink detection models are not based on Greek letters, optical character recognition (OCR), or language models. Instead, they independently detect tiny spots of ink in the CT scan, the writing appearing later when these are aggregated. As a result, the text appearing in the images is not the imagined output of a machine learning model, but is instead directly tied to the underlying data in the CT scan. […] This makes it extremely unlikely for the model to hallucinate shapes that look like letters.

    https://scrollprize.org/grandprize section “How accurate are those pictures” avec une super animation qui explique très bien la technique et pourquoi elle ne va pas inventer de texte.

    Elle ne remplit pas de trou dans les textes eux même.