1. SINetV2[19], BGNet [15]) are built on Res2Net-50 [69], we implement TSNet+ using Res2Net-50 as the backbone for fair comparison. Quantitative Evaluation. The quantitative evaluation results of all models are shown in Table 1. It can be clearly seen from the table that TSNet surpasses other high-performance models across all benchmark datasets in terms of all evaluation metrics. More specifically, performance gains over the three best compared algorithms;Fnet;Since some competing methods