DREAM introduces a two-stage adaptive multi-modal fusion framework that reaches BLEU-4 of 0.241 on DeepEyeNet for retinal image report generation and generalizes to ROCO.
LLaMA: Open and efficient foundation language models
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
-
DREAM: Dynamic Retinal Enhancement with Adaptive Multi-modal Fusion for Expert Precision Medical Report Generation
DREAM introduces a two-stage adaptive multi-modal fusion framework that reaches BLEU-4 of 0.241 on DeepEyeNet for retinal image report generation and generalizes to ROCO.