LRANet++: Low-Rank Approximation Network for Accurate and Efficient Text Spotting

Benchmark Model Rank Results
text-spotting-on-scut-ctw1500LRANet++ (Synth150K, MLT17, Total-Text, IC13, IC15, TextOCR)#1F-measure (%) - No Lexicon: 70.7F-Measure (%) - Full Lexicon: 85.2
text-spotting-on-scut-ctw1500LRANet++ (Synth150K, MLT17, Total-Text)#2F-measure (%) - No Lexicon: 67.7F-Measure (%) - Full Lexicon: 83.8
text-spotting-on-total-textLRANet++ (Synth150K, MLT17, IC13, IC15, TextOCR)#1F-measure (%) - No Lexicon: 84.6F-measure (%) - Full Lexicon: 89.7
text-spotting-on-total-textLRANet++ (Synth150K, MLT17)#4F-measure (%) - No Lexicon: 82.2F-measure (%) - Full Lexicon: 87.7
text-spotting-on-icdar-2015LRANet++ (Synth150K, MLT17, Total-Text, IC13, IC15, TextOCR)#3F-measure (%) - Strong Lexicon: 88.1F-measure (%) - Weak Lexicon: 84.0…
text-spotting-on-icdar-2015LRANet++ (Synth150K, MLT17, Total-Text)#5F-measure (%) - Strong Lexicon: 87.0F-measure (%) - Weak Lexicon: 82.5…