EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks

RT-1-X uses EfficientNet + FiLM.