Supplementary Materials for Disentangled High Quality Salient Object Detection

*Corresponding author and equal contribution to first author. This work was supported by National Natural Science Foundation of China 61906036 and the Fundamental Research Funds for the Central Universities (2242021k30056). methods across these two conventional datasets. We also show their PR curves in Fig.1. It should be noted that Fmax represents F β . We apologize for this writing error of Table.2 in the main text. F-measure curves of different methods are displayed in Fig.2, for overall comparisons. One can observe that our approach noticeably outperforms all the other state-of-theart methods. These observations demonstrate the efficiency and robustness of our proposed method across various challenging datasets. SOC [1] is a new challenging dataset with nine attributes. In Table.2, we evaluate the mean F-measure score of our method as well as 11 state-of-the-art methods. We can see the proposed model achieves the competitive results among most of attributes and the overall score is best. Model size and running time comparisons among different methods are also reported in Table.3. It can be seen that with the high-resolution input, our method is more efficient than HRNet. For fair, the running time analysis of our method is also conducted with the low-resolution input (352×352), and our method runs at a competitive efficiency.