List your company
ResearchFormerly WizWand

Document Classification on RVL-CDIP (test)

97.7AccuracyEAML_Tr-KLD_Reg
94.299295.182196.06596.9479Dec 29, 2020Aug 27, 2021Apr 26, 2022Dec 23, 2022Aug 22, 2023Apr 19, 2024Dec 17, 2024
Updated 8mo ago

Evaluation results

MethodLinks
EAML_Tr-KLD_RegModality=Multi-modal2023.0597.7
EAML_Tr-KLD_RegModality=Image2023.0597.67
EAML_Tr-KLD_RegModality=Text2023.0597.63
Cross-ModalModality=Multi-modal2023.0597.05
ERNIE-Layout_largeVariant=Large2022.1096.27
ERNIE-LayoutModality=T+L+I, Model size=Large2023.0596.27
UDOP-DualModality=V+T+L2022.1296.22
DocFormer_BASEuse OCR=true, # params=183M + OCR2022.0396.2
DocFormer_BASE2022.0296.17
DocFormerModel scale=base2022.1096.17
DocFormer BaseModalities=V + T, #Param.=183M2023.0396.17
DocFormer B aseModality=Vision + Language + Layout, Train Data=5M, #Params=183M2022.0596.17
DocFormerModality=T+L+I, Model size=Base2023.0596.17
DocFormer BaseModality=Vision + Text + Layout, Pretrain Data=5M2023.0996.17
DocFormerCategory=Text-Based Methods, #Param=183M2024.1296.17
DocFormer-base#param (M)=183, Input modalities=image + text + spatial2021.0696.17
GraphDocResnetModality=Vision + Text + Layout, Pretrain Data=320k2023.0996.1
StructuralLM_LARGEParams=355M2021.0596.08
StructuralLM_largeVariant=Large2022.1096.08
StructuralLM LargeModalities=T, #Param.=355M2023.0396.08
StructuralLMlargeModality=Text+Layout2022.1296.08
StructuralLMlargeModality=T+L2022.1296.08
StructuralLMModality=T+L, Model size=Large2023.0596.08
GraphDocModality=Vision + Text + Layout, Pretrain Data=320k2023.0996.02
UDOPModality=Vision+Text+Layout2022.1296
UDOPModality=V+T+L2022.1296
LayoutLMv3 LargeModalities=V + T, #Param.=368M2023.0395.93
LayoutLMv3largeModality=Vision+Text+Layout2022.1295.93
LayoutLMv3largeModality=V+T+L2022.1295.93
LayoutLMv3Modality=T+L+I, Model size=Large2023.0595.93
ERNIE-LayoutModel scale=base2022.1095.81
LILT[EN-R]_BASEtext_backbone=RoBERTa-base2022.0295.68
LILTModality=Text+Layout2022.1295.68
LILTModality=T+L2022.1295.68
LiLTModality=T+L+I, Model size=Base2023.0595.68
LiLT BaseModality=Text + Layout, Pretrain Data=11M2023.0995.68
LayoutLMv2-large#param (M)=426, Input modalities=image + text + spatial2021.0695.65
LayoutLMv2Size variant=Large, Parameters=426M2021.0295.64
LayoutLMv2-LARGEInput=Multi-modal2020.1295.64
LayoutLMv2-LargePretraining Source=IIT-CDIP, Pretraining #Data=11M, Pretraining Scale=Word, Pretraining Max #Words=512, Pretraining Modality=V+L, #Param.=426M2022.0495.64
LayoutLMv2_largeVariant=Large2022.1095.64
LayoutLMv2 LargeModalities=V + T, #Param.=426M2023.0395.64
LayoutLMv2largeModality=Vision+Text+Layout2022.1295.64
LayoutLMv2largeModality=V+T+L2022.1295.64
Xu et al.2023.0595.64
LayoutLMv2Modality=T+L+I, Model size=Large2023.0595.64
LiLT[InfoXLM]_BASEtext_backbone=InfoXLM-base, multilingual=true2022.0295.62
BROS_BASE2022.0295.58
BROSModality=Text + Layout, Pretrain Data=11M2023.0995.58
BROS-base#param (M)=139, Input modalities=text / (text + spatial)2021.0695.58
TILTSize variant=Large, Parameters=780M2021.0295.52
TILT-LargePretraining Source=RVL-CDIP+, Pretraining #Data=1.1M, Pretraining Scale=Word, Pretraining Max #Words=512, Pretraining Modality=V+L, #Param.=780M2022.0495.52
TILT_largeVariant=Large2022.1095.52
TILT LargeModalities=V + T, #Param.=780M2023.0395.52
TILTlargeModality=Vision+Text+Layout2022.1295.52
TILTlargeModality=V+T+L2022.1295.52
TITLModality=T+L+I, Model size=Large2023.0595.52
DocFormer_largeVariant=Large2022.1095.5
DocFormerlargeModality=Vision+Text+Layout2022.1295.5
DocFormerlargeModality=V+T+L2022.1295.5
DocFormerModality=T+L+I, Model size=Large2023.0595.5
DocFormer-large#param (M)=536, Input modalities=image + text + spatial2021.0695.5
LayoutLMv3 B aseModality=Vision + Language + Layout, Train Data=11M, #Params=133M2022.0595.44
LayoutLMv3Modality=T+L+I, Model size=Base2023.0595.44
LayoutLMv3 BaseModality=Vision + Text + Layout, Pretrain Data=11M2023.0995.44
LayoutLMv3Category=Text-Based Methods, #Param=133M2024.1295.44
LayoutLMv2_BASEuse OCR=true, # params=200M + OCR2022.0395.3
DonutOCR=false, #Params=143M, Time (ms)=7522021.1195.3
DonutModality=Vision2022.1295.3
DonutModality=V2022.1295.3
LayoutLMv2Size variant=Base, Parameters=200M2021.0295.25
TILTSize variant=Base, Parameters=230M2021.0295.25
LayoutLMv2-BASEInput=Multi-modal2020.1295.25
LayoutLMv2_BASE2022.0295.25
LayoutLMv2-BasePretraining Source=IIT-CDIP, Pretraining #Data=11M, Pretraining Scale=Word, Pretraining Max #Words=512, Pretraining Modality=V+L, #Param.=200M2022.0495.25
TILT-BasePretraining Source=RVL-CDIP+, Pretraining #Data=1.1M, Pretraining Scale=Word, Pretraining Max #Words=512, Pretraining Modality=V+L, #Param.=230M2022.0495.25
LayoutLMv2OCR=true, #Params=200M + α†, Time (ms)=14892021.1195.25
TILTModel scale=base2022.1095.25
LayoutLMv2Model scale=base2022.1095.25
TIL T B aseModality=Vision + Language + Layout, Train Data=1M, #Params=230M2022.0595.25
LayoutLMv2 B aseModality=Vision + Language + Layout, Train Data=11M, #Params=200M2022.0595.25
TITLModality=T+L+I, Model size=Base2023.0595.25
LayoutLMv2Modality=T+L+I, Model size=Base2023.0595.25
TILT BaseModality=Vision + Text + Layout, Pretrain Data=1M2023.0995.25
LayoutLMv2 BaseModality=Vision + Text + Layout, Pretrain Data=11M2023.0995.25
LayoutLMv2-base#param (M)=200, Input modalities=image + text + spatial2021.0695.25
LayoutXLM_BASEmultilingual=true2022.0295.21
UDoc*Pretraining Source=IIT-CDIP, Pretraining #Data=1M, Pretraining Scale=Region, Pretraining Max #Words=64x512, Pretraining Modality=V+L, #Param.=272M, OCR Engine=Google OCR2022.0495.05
UDocModalities=V + T, #Param.=272M2023.0395.05
UniDocModality=Vision+Text+Layout2022.1295.05
UDocModality=V+T+L2022.1295.05
UDocModality=Vision + Text + Layout, Pretrain Data=11M2023.0995.05
StrucTexTv2 LargeModalities=V, #Param.=238M2023.0394.62
Donutuse OCR=false, # params=156M2022.0394.5
EAM LModality=Vision + Language, Train Data=320k2022.0594.44
LayoutLM_LARGEParams=390M2021.0594.43
LayoutLMSize variant=Large, Parameters=343M2021.0294.43
LayoutLM-LARGEInput=w/ image2020.1294.43
LayoutLM-LargePretraining Source=IIT-CDIP, Pretraining #Data=11M, Pretraining Scale=Word, Pretraining Max #Words=512, Pretraining Modality=L, #Param.=343M2022.0494.43
LayoutLM_largeVariant=Large2022.1094.43