Table 2.
Ablation study.
Datasets | NJUD | NLPR | STERE | SIP | ||||||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
Metrics | MAE | FM | SM | EM | MAE | FM | SM | EM | MAE | FM | SM | EM | MAE | FM | SM | EM |
① | 0.041 | 0.910 | 0.899 | 0.906 | 0.023 | 0.920 | 0.921 | 0.957 | 0.035 | 0.911 | 0.905 | 0.925 | 0.049 | 0.900 | 0.879 | 0.918 |
② | 0.041 | 0.909 | 0.901 | 0.906 | 0.026 | 0.91 | 0.916 | 0.948 | 0.039 | 0.903 | 0.9 | 0.915 | 0.051 | 0.894 | 0.878 | 0.916 |
③ | 0.046 | 0.907 | 0.895 | 0.916 | 0.031 | 0.903 | 0.906 | 0.941 | 0.073 | 0.86 | 0.82 | 0.869 | 0.056 | 0.897 | 0.869 | 0.907 |
④ | 0.040 | 0.911 | 0.901 | 0.936 | 0.022 | 0.922 | 0.923 | 0.957 | 0.035 | 0.912 | 0.903 | 0.937 | 0.047 | 0.902 | 0.883 | 0.925 |
Ours | 0.039 | 0.914 | 0.902 | 0.936 | 0.022 | 0.922 | 0.924 | 0.958 | 0.036 | 0.914 | 0.905 | 0.941 | 0.046 | 0.904 | 0.886 | 0.926 |
The ‘ours’ in Table 2 means our proposed method. The ① refers to the experimental results by replacing the structure of FiCaps with U-Nets. The ② means the experimental results, which FiCaps does not integrate with external features. The ③ indicates the experimental results, extracting and integrating the features of depth image by using the VGG backbone. The ④ refers to the experimental results by replacing the GCM with the traditional convolutions.