A text-anchored distillation loss plus ImageNet predistillation lets small vision-language student models match a larger ResNet-50 teacher on placenta pathology tasks while running several times faster.
In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
VLCD: Vision-Language Contrastive Distillation for Accurate and Efficient Automatic Placenta Analysis
A text-anchored distillation loss plus ImageNet predistillation lets small vision-language student models match a larger ResNet-50 teacher on placenta pathology tasks while running several times faster.