chunked/nltk_chunking/_tasks_explained/chunk_109.txt

Like DETR, Mask2Former also uses learned object queries and combines them with the image features from the pixel decoder to make a set prediction (class label, mask prediction).