Cross-Domain Isolated Sign Language Recognition Setting

Cross-Camera ISLR aims to test the robustness of the model against variations in camera specifications and settings. Training and testing data are captured from different cameras. It is challenging for the model to generalize across hardware-induced discrepancies.

Cross-View ISLR requires the model to recognize signs from views not seen during training. With training views denoted as $V_{\text{train}}$ and testing views as $V_{\text{test}}$, the model must handle the appearance changes due to different viewing angles, thus testing its view-invariance capabilities.