Frame-Level Real-Time Assessment of Stroke Rehabilitation Exercises from Video-Level Labeled Data: Task-Specific vs. Foundation Models

Gonçalo Mesquita, Ana Rita Cóias, Artur Dubrawski, and Alexandre Bernardino · 2025

Read Frame-Level Real-Time Assessment of Stroke Rehabilitation Exercises from Video-Level Labeled Data: Task-Specific vs. Foundation Models by Gonçalo Mesquita, Ana Rita Cóias, Artur Dubrawski, and Alexandre Bernardino (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2506-03752.txt.

Description

Gonçalo Mesquita and co-authors tackle frame-level scoring of stroke rehabilitation exercises, in a 2025 paper. Patients who practice alone at home need fast feedback on their movements. Virtual coaches can watch video and flag poor form in real time. Training such systems needs labels for each video frame, which are slow and costly to make.

The framework learns from labels given only for whole videos. A gradient-based method plus careful pseudo-label choice creates frame labels for training. The team compares a task-specific model with large foundation models. Their tests show which route works best for spotting compensatory motions. The aim is safe, low-cost coaching for home recovery.

Catalog record

CREATORGonçalo Mesquita, Ana Rita Cóias, Artur Dubrawski, and Alexandre Bernardino
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE30 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2506-03752

Browse the full library · About the Tower of Babel