TEACH: Text Encoding as Curriculum Hints for Scene Text Recognition
This addresses STR challenges like complex visual appearances for applications in document analysis and image understanding, but it is incremental as it builds on existing encoder-decoder frameworks.
The authors tackled the problem of Scene Text Recognition (STR) by proposing TEACH, a training paradigm that injects ground-truth text as auxiliary input and progressively reduces its influence, leading to consistently improved accuracy across multiple benchmarks.
Scene Text Recognition (STR) remains a challenging task due to complex visual appearances and limited semantic priors. We propose TEACH, a novel training paradigm that injects ground-truth text into the model as auxiliary input and progressively reduces its influence during training. By encoding target labels into the embedding space and applying loss-aware masking, TEACH simulates a curriculum learning process that guides the model from label-dependent learning to fully visual recognition. Unlike language model-based approaches, TEACH requires no external pretraining and introduces no inference overhead. It is model-agnostic and can be seamlessly integrated into existing encoder-decoder frameworks. Extensive experiments across multiple public benchmarks show that models trained with TEACH achieve consistently improved accuracy, especially under challenging conditions, validating its robustness and general applicability.