TILA uses temporal inversion of image pairs as a supervisory signal to make existing temporal vision-language models more sensitive to directional interval changes in chest X-rays.
Projection layers for both image and text en- coders have dimension 128, and the CXR-BERT text en- coder is configured with a maximum token length of 256
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Temporal Inversion for Learning Interval Change in Chest X-Rays
TILA uses temporal inversion of image pairs as a supervisory signal to make existing temporal vision-language models more sensitive to directional interval changes in chest X-rays.