Transferable Feature Representation for Visible-to-Infrared Cross-Dataset Human Action Recognition

Figure 8

The comparison of the confusion matrices between the baseline method and the proposed method in InfAR dataset, where the left is from the baseline method, while the right is from our proposed CDFAG. Note that “hc” denotes “handclapping” and “hs” denotes “handshaking,” respectively.

(a) Confusion matrix of the baseline method

(b) Confusion matrix of the proposed method