What Is Wrong with Synthetic Data for Scene Text Recognition? A Strong Synthetic Engine with Diverse Simulations and Self-Evolution

Benchmark Model Rank Results
scene-text-recognition-on-svtSVTRv2-AR + UnionST-SP + ISR (D_P2 + D_L-hard)#1Accuracy: 99.23
scene-text-recognition-on-union14m-benchmarkSVTRv2-AR + UnionST-SP + R#1WAICS (Curve): 95.42WAICS (Artistic): 86.67WAICS (Salient): 91.09…
scene-text-recognition-on-union14m-benchmarkSVTRv2-AR + UnionST-PR + R#9WAICS (Multi-Oriented): 97.3
scene-text-recognition-on-union14m-benchmarkSVTRv2-AR + UnionST-SP + ISR (D_P2 + D_L-hard)#10WAICS (Contextless): 91.14
scene-text-recognition-on-union14m-benchmarkSVTRv2-AR + UnionST-SR + R#11WAICS (Multi-Words): 92.48
scene-text-recognition-on-svtpMAERec + UnionST-S#3Accuracy: 98.14
scene-text-recognition-on-iiit5kSVTRv2-AR + UnionST-SP + R#6Accuracy: 99.47