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

Table 4

Parameter comparison in ablation models during training and testing (M).

Model	Parameters
	Training	Testing
Baseline (HRNet)	35.564	35.564
Baseline + SR (Bilinear Interpolation)	35.565	35.564
Baseline + SR (Bilinear Interpolation) + SAM	35.568	35.567
Baseline + SR (Trans Conv2D) + SAM	35.573	35.567
Proposed model (Baseline + SR (Trans Conv2D) + SAM + Independent Trans Conv2D Layers)	37.582	37.576