Table 3 Model complexity and speed evaluation of different methods.
From: Lightweight monocular depth estimation using a fusion-improved transformer
Methods | Params (M) | FLOPs (G) | RMSE | \(\delta\)< 1.25 | FPS | ||
|---|---|---|---|---|---|---|---|
Raspberry Pi | SBC-T800 | Titan X FPS | |||||
FastDepth | 3.96 | 3.6 | 5.321 | 0.808 | 11.39 | 57.68 | 116.20 |
Monodepth2 | 14.3 | 8.3 | 5.927 | 0.877 | 9.96 | 47.16 | 90.37 |
R-MSFM6 | 3.8 | 31.2 | 4.704 | 0.876 | 5.87 | 19.87 | 43.84 |
Lite-Mono | 3.1 | 5.1 | 4.561 | 0.886 | 7.66 | 35.45 | 67.92 |
Ours | 3.0 | 4.6 | 4.506 | 0.891 | 8.79 | 40.95 | 87.71 |
