Table 8.
Computation time (milliseconds per frame) for various deep learning models under different hardware configurations. “CPU” implies using a single CPU, “CPU multi-thread” indicates performing multi-threaded processing, “CPU + GPU” denotes using the GPU with a single CPU thread, and “CPU multi-thread + GPU” represents using the full capability of the computer system discussed in Section 2.4.
| CPU Only | CPU Multi-Thread | CPU + GPU | CPU Multi-Thread + GPU | |
|---|---|---|---|---|
| Unet++ | 632.1 | 624.1 | 26.7 | 17.0 |
| CaraNet | 521.3 | 506.7 | 76.1 | 66.8 |
| SSFormer-S | 441.8 | 436.6 | 44.4 | 32.0 |
| ESFPNet-T | 144.2 | 135.0 | 34.6 | 20.7 |
| ESFPNet-S | 369.1 | 360.1 | 44.4 | 31.9 |
| ESFPNet-L | 759.9 | 754.1 | 88.2 | 73.6 |