Codesota · Models · DiT-L (Cascade R-CNN)Microsoft Research1 results · 1 benchmarks

Model card

DiT-L (Cascade R-CNN).

Microsoft Researchopen-sourceUnknown paramsDocument Image Transformer (BEiT-based) + Cascade R-CNN detection head

DiT: Self-supervised Pre-Training for Document Image Transformer. Large variant (307M params) fine-tuned with Cascade R-CNN head. SOTA on DocLayNet at time of publication. CVPR 2022 Workshop. arXiv 2203.02155.

§ 01 · Benchmarks

Every benchmark DiT-L (Cascade R-CNN) has a recorded score for.

#	Benchmark	Area · Task	Metric	Value	Rank	Date	Source
01	DocLayNet	Computer Vision · Document Understanding	mAP	82.6%	#2/7	—	source ↗

Rank column shows this model’s position vs all other models scored on the same benchmark + metric (competitors after the slash). #1 in red means current SOTA. Sorted by rank, then newest result.

§ 02 · Strengths by area