A CNN-GAT-ViT hybrid that patches feature maps and uses graph attention; reported gains are inflated by near-random self-made baselines and internal inconsistencies.
Uni4eye++: A general masked image modeling multi-modal pre-training framework for ophthalmic image classification and segmentation
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2024 1verdicts
REJECT 1representative citing papers
citing papers explorer
-
SAG-ViT: A Scale-Aware, High-Fidelity Patching Approach with Graph Attention for Vision Transformers
A CNN-GAT-ViT hybrid that patches feature maps and uses graph attention; reported gains are inflated by near-random self-made baselines and internal inconsistencies.