Sample RankingRecallPrecisionHmean
DenseNet-Attention Recognition96.40%95.54%95.96%
BOE_IOT_AIBD93.69%93.69%93.69%
HeReceipt-Ensemble92.79%93.64%93.21%
H&H Lab92.79%92.79%92.79%
Unet followed by CRNN with CTC92.79%91.96%92.38%
Ping An Property & Casualty Insurance Company90.09%91.74%90.91%
CRNN after UNet Segmentation91.89%88.70%90.27%
CLOVA OCR90.99%88.60%89.78%
SCUT-DLVC-Lab-Lexicon90.09%87.72%88.89%
CITlab Argus Text Recognition87.39%85.09%86.22%
IFLYTEK-textRec_v485.59%86.36%85.97%
Attention based OCR85.59%78.51%81.90%
SituTech_OCR79.28%80.00%79.64%
BiLstmCtcIgnoreSpaces-Segment73.87%73.21%73.54%
CTPN_CRNN32.43%58.06%41.62%
A Text Extraction Method Based on Modified CRNN27.93%75.61%40.79%
BiLSTM+ctc25.23%41.18%31.28%
CRNN_Pytorch_BiCTC14.41%34.78%20.38%
ICA-IVA13.51%31.25%18.87%
Receipt Info Extracting Task2 letter-recognizing0.90%3.33%1.42%
VIL0.00%0.00%0.00%
Task 2 - Scanned Receipt OCR (Submitted by Intuit Inc.)0.00%0.00%0.00%
CRNN_Autohome0.00%0.00%0.00%