Alapfogalmak
Document Layout Analysis models' robustness is benchmarked using RoDLA, introducing a taxonomy of perturbations and proposing metrics for evaluation.
Kivonat
The content introduces RoDLA, a benchmark for assessing the robustness of Document Layout Analysis models. It covers the taxonomy of document perturbations, proposed metrics like mPE and mRD, and introduces the RoDLA model with experiments on three datasets. The study highlights the importance of evaluating DLA models' robustness in real-world scenarios.
- Introduction to Document Layout Analysis challenges.
- Proposal of a robustness benchmark for DLA models.
- Taxonomy of document perturbations and severity levels.
- Introduction of metrics like mPE and mRD for evaluation.
- Description of the RoDLA model with experiments on datasets.
- Comparison with state-of-the-art methods in terms of performance and robustness.
Statisztikák
450K document images used for benchmarking.
RoDLA achieves state-of-the-art mAP scores on clean datasets.
Perturbation taxonomy includes 12 types inspired by real-world processing.
Idézetek
"Conducting comprehensive robustness testing is essential before developing a DLA model."
"Our RoDLA method improves attention mechanisms to extract robust features."