Table 10 Performance comparison of individual models and the weight-centric mechanism based on computational resources.

From: MythicVision: a deep learning powered mobile application for understanding Indian mythological deities using weight centric decision approach

Employed deep neural networks

No. of parameters

GFLops

Accuracy (%)

Average Inference time (in milli-seconds)

MobileNetV2

 ~ 3 M

0.43

93

12.68

Efficient Net B0

 ~ 5 M

0.59

95

22.21

ResNet-50

 ~ 25 M

7.72

93

25.96

GoogleNet (Inception V3)

 ~ 22 M

5.68

92

30.37

Weight-centric decision mechanism

54 M

14.42

96

55.61

Back to article page