Fig. 4From: PSENet-based efficient scene text detectionThe ResNet is optimized by combined ResNeXt and Res2Net (Variable cardinality C = 32, scale dimension s = 4)Back to article page