2d rotary embedding 1anisotropic layouts 1cross-attention 1positional encoding 1scene text recognition 1transformer models 1
From the 1 of 2 linked papers with an AI index.
2 papers
cs.CV2026
Out-of-Length Scene Text Recognition: A Two-Axis Diagnosis and a Training-Free Fix
Zobeir Raisi, John Zelek
Scene Text Recognition (STR) models are trained almost exclusively on word crops of at most 25 characters, yet real deployments (signage, product labels, dense captions) require re…
cs.CV2026
2D Rotary Position Embedding for Scene Text Recognition with Transformers
Zobeir Raisi
The paper introduces a parameter‑free 2D rotary position embedding (2D‑RoPE) for transformer‑based scene text recognition, adapting axial RoPE to anisotropic text images and extend…