Skip to main content
. 2023 Jul 13;23(14):6384. doi: 10.3390/s23146384

Table 5.

Comparing self-supervised pre-trained models for HAR using HMDB51.

Method Accuracy (%)
Order Prediction Network pre-trained w/ UCF101 [77] 37.5
Odd-one-out pre-trained w/ ImageNet [76] 40.8
2D ViT + LSTM pre-trained w/ ImageNet (Present study) 41.9