Table 5 Inference times in seconds from the experiments on Jetson Orin Nano using the tiny versions of the networks.
From: Efficient attention vision transformers for monocular depth estimation on resource-limited hardware
Model | Jetson NYU [s] \(\downarrow\) | Jetson KITTI [s] \(\downarrow\) |
|---|---|---|
METER | 328.49 | 343.79 |
Meta METER | 215.33 | 221.21 |
Pyra METER | 310.59 | 336.41 |
MoH METER | 344.43 | 351.07 |
PXF | 1924.38 | 2765.59 |
Meta PXF | 1373.43 | 1970.51 |
Meta-Base PXF | 1543.57 | 2216.83 |
Base-Meta PXF | 1722.62 | 2469.82 |
Pyra PXF | 1915.70 | 2709.43 |
Pyra-Base PXF | 1927.62 | 2712.86 |
Base-Pyra PXF | 1906.42 | 2735.54 |
MoH PXF | 2027.07 | 2876.13 |
MoH-Base PXF | 2000.42 | 2844.66 |
Base-MoH PXF | 1942.72 | 2809.53 |
NeWCRFs | 2316.55 | 3185.19 |
Meta NeWCRFs | 1615.87 | 2235.18 |
Meta-Base NeWCRFs | 1958.86 | 2730.05 |
Base-Meta NeWCRFs | 2007.11 | 2779.84 |
Pyra NeWCRFs | 2380.42 | 3391.28 |
Pyra-Base NeWCRFs | 2323.66 | 3241.11 |
Base-Pyra NeWCRFs | 2368.85 | 3281.75 |
MoH NeWCRFs | 2482.51 | 3456.72 |
MoH-Base NeWCRFs | 2375.26 | 3329.97 |
Base-MoH NeWCRFs | 2292.42 | 3409.21 |
