Inverting Visual Representations with Detection Transformers