DoPTA: Improving Document Layout Analysis using Patch-Text Alignment

Benchmark Model Rank Results
document-image-classification-on-rvl-cdipDoPTAAccuracy: 94.12%Parameters: 85M
document-layout-analysis-on-d4laDoPTAmAP: 70.72Model Parameters: 85M
document-layout-analysis-on-publaynet-valDoPTAOverall: 0.949Text: 0.944Title: 0.895List: 0.957Table: 0.977