HRNet Encoder and Dual-Branch Decoder Framework-Based Scene Text Recognition Model

Table 3

PSNR results of restored images by Trans Conv2D and bilinear interpolation (dB).

	Bilinear interpolation	Trans Conv2D	Improved
IIIT5k	27.57	30.88	+3.31
SVT	31.93	36.05	+4.12
IC03	27.79	31.68	+3.89
IC13	27.82	32.06	+4.24
IC15	33.08	38.11	+5.03
SVTP	32.76	37.86	+5.10
CUTE80	24.98	28.39	+3.41