Table 3 Comparison of computational cost on different models.

From: Frame topology fusion-based hierarchical graph convolution for automatic assessment of physical rehabilitation exercises

Phase

Methods

Num. of videos

# Params \(\downarrow\)

FLOPs \(\downarrow\)

Exe. time \(\downarrow\)

Avg. Time \(\downarrow\)

MAD \(\downarrow\)

RMSE \(\downarrow\)

MAPE \(\downarrow\)

Train

Deb et al.12

373

0.722M

26B

3369s

Mourchid et al.13

373

0.572M

19B

1574s

Ours

373

0.607M

21B

2169s

Test

Deb et al.12

100

0.722M

26B

13.16s

131ms

0.621

1.181

1.591

Mourchid et al.13

100

0.572M

19B

3.06s

31ms

0.399

0.735

1.217

Ours

100

0.607M

21B

5.31s

53ms

0.390

0.698

1.113

  1. The bolded text indicates the best results.