Table 4 Model architecture specifications with computational metrics and execution time. It is to be noted that “Per-step” (in ms) and “Total time” (in s) indicate the average time taken to process a single batch during inference and the total inference time for the complete test dataset, respectively.
From: JUHCCR-v1: a database for hand-drawn electrical and electronics circuit component recognition
Model | # Parameters | GFLOPs | Execution time | |||
|---|---|---|---|---|---|---|
Trainable | Non-trainable | Total | Per-step (ms) | Total (s) | ||
MobileNet-V2 | 25,620 | 2,257,984 | 2,283,604 | 0.3 | 887 | 30 |
NasNet-Mobile | 21,140 | 4,269,716 | 4,290,856 | 1.1 | 1000 | 43 |
ResNet-50 | 40,980 | 23,587,712 | 23,628,692 | 4.1 | 3000 | 108 |
Inception-V3 | 40,980 | 21,802,784 | 21,843,764 | 6.0 | 2000 | 53 |
DenseNet-121 | 20,500 | 7,037,504 | 7,058,004 | 2.9 | 151 | 8 |
Present | 282,743 | 7,037,504 | 7,320,247 | 3.3 | 4000 | 139 |
