Table 2 Ablation study of the proposed method. Baseline denotes the results of 2D methods over all-focus images while Avg. denotes the average results of 2D methods over all slices. Img. Input and Sal. Input denote the model trained only with RGB images and saliency maps, respectively. Reg. Loss and w/o Com. denote learning the model with the regression loss and without the proposed comparator, respectively. Bub. Test denotes testing the trained model as a bubble sorting algorithm. R-18 and SER-18 denote replacing the proposed encoder with well-designed architecture ResNet-18 [50] and SE-ResNet-18 [51], respectively
From: Are RGB-based salient object detection methods unsuitable for light field data?
Dataset | Variant | GCPA | Â | F3Net | Â | SCRN | Â | BASNet | CPD | Â | |
|---|---|---|---|---|---|---|---|---|---|---|---|
| Â | Â | \(\boldsymbol{\mathcal {M}}\downarrow \) | \(\boldsymbol{\mathcal {E}}\uparrow \) | \(\boldsymbol{\mathcal {M}}\downarrow \) | \(\boldsymbol{\mathcal {E}}\uparrow \) | \(\boldsymbol{\mathcal {M}}\downarrow \) | \(\boldsymbol{\mathcal {E}}\uparrow \) | \(\boldsymbol{\mathcal {M}}\downarrow \) | \(\boldsymbol{\mathcal {E}}\uparrow \) | \(\boldsymbol{\mathcal {M}}\downarrow \) | \(\boldsymbol{\mathcal {E}}\uparrow \) |
DUTLF | Ours | .045 | .928 | .045 | .926 | .048 | .917 | .048 | .924 | .046 | .922 |
| Â | Baseline | .076 | .874 | .070 | .878 | .074 | .868 | .064 | .894 | .067 | .885 |
| Â | Avg. | .085 | .851 | .084 | .847 | .088 | .838 | .082 | .873 | .084 | .869 |
| Â | Img. Input | .054 | .903 | .056 | .892 | .061 | .881 | .058 | .900 | .055 | .897 |
| Â | Sal. Input | .046 | .926 | .049 | .920 | .500 | .913 | .053 | .918 | .050 | .914 |
| Â | Reg. Loss | .050 | .919 | .050 | .917 | .053 | .910 | .052 | .918 | .053 | .914 |
| Â | w/o Com. | .049 | .921 | .053 | .914 | .053 | .908 | .057 | .910 | .054 | .911 |
| Â | Bub. Test | .048 | .923 | .051 | .915 | .050 | .914 | .049 | .923 | .049 | .914 |
| Â | R-18 | .049 | .919 | .047 | .923 | .051 | .916 | .054 | .911 | .048 | .916 |
| Â | SER-18 | .048 | .922 | .047 | .922 | .049 | .920 | .051 | .917 | .048 | .916 |
LFSD | Ours | .070 | .882 | .070 | .892 | .066 | .900 | .078 | .888 | .071 | .892 |
| Â | Baseline | .098 | .835 | .105 | .830 | .087 | .861 | .084 | .870 | .089 | .860 |
| Â | Avg. | .105 | .829 | .108 | .820 | .096 | .843 | .089 | .857 | .101 | .839 |
| Â | Img. Input | .072 | .882 | .072 | .878 | .073 | .885 | .080 | .878 | .079 | .882 |
| Â | Sal. Input | .070 | .881 | .073 | .884 | .067 | .895 | .087 | .872 | .074 | .888 |
| Â | Reg. Loss | .076 | .874 | .079 | .876 | .070 | .893 | .080 | .884 | .074 | .890 |
| Â | w/o Com. | .078 | .871 | .075 | .884 | .065 | .897 | .086 | .873 | .071 | .891 |
| Â | Bub. Test | .076 | .875 | .081 | .874 | .066 | .896 | .079 | .887 | .073 | .892 |
| Â | R-18 | .077 | .875 | .072 | .884 | .067 | .892 | .079 | 884 | .075 | .881 |
| Â | SER-18 | .076 | .875 | .073 | .886 | .066 | .896 | .079 | .885 | .073 | .887 |