method: Res2Net2020-01-08

Authors: suomi

Description: res2net50_26w_4s
no data augmentation
use only pure IC17 training set (without validation set)

method: Multi_scale_v22019-09-27

Authors: Wuheng Xu, Changxu Cheng, Bohan Li

Description: Area block feature information using images on multiple scales.This model has 4 scales and 8 branches.We used three training sets(mlt17, mlt19, mlt19val).

method: 4Paradigm-Data-Intelligence2019-05-30

Authors: ACVG

Description: Recognition model: Based on Transformer with backbone ResNet50. A voting process is done to identify the language of recognized transcript. Train-set: 2017 MLT task2 train-set & 2019 MLT task2 train-set & 2019 MLT Synthetic dataset.

Ranking Table

Description Paper Source Code
DateMethodScript classification accuracy
2020-01-08Res2Net90.69%
2019-09-27Multi_scale_v290.14%
2019-05-304Paradigm-Data-Intelligence90.12%
2020-01-10RAMSI89.66%
2019-09-27Convolution based method 389.60%
2019-09-26Convolution based method 289.43%
2019-01-19HUST(GSPA)89.42%
2018-08-01CLOVA-AI / PAPAGO89.01%
2019-01-17HUST(GS)88.51%
2019-09-10CNN based method88.36%
2017-07-02CNN based method 788.09%
2017-06-28SCUT-DLVClab87.69%
2017-07-01CNN based method 487.33%
2017-07-01CNN based method 586.97%
2017-06-30CNN based method 286.60%
2017-07-02BLCT86.34%
2017-07-02BLCT86.24%
2017-06-02ecn-based method82.20%
2017-07-01TH-DL80.72%
2017-07-01An approach towards Word-Level Multi-Script Identification using Deep Transfer Features and SVM74.81%
2017-07-01TNet48.33%
2017-07-01TH-CNN43.22%

Ranking Graphic