About parameters setting for ablation experiments

[babble2]

Excuse me，when performing ablation experiments in table 2 of the paper, are the parameters set correctly?

[retsuh-bqw]

Please refer to configs/SRFormer/TotalText/R_50_poly.yaml, and use (totaltext_poly_train_rotate_pos) as the training data when performing Tab.2.
In addition, if you have limited computing resources and can only use BatchSize=1, you need to consider rescaling the learning rate, to avoid the gradient stochasticity impact on the optimization (but there is no guarantee that you can reproduce the results under bs=8).