GabriellaV2: Towards better generalization in surveillance videos for Action Detection | IEEE Conference Publication | IEEE Xplore