raw)

        >>> image_processor = AutoImageProcessor.from_pretrained("facebook/detr-resnet-50")
        >>> model = DetrModel.from_pretrained("facebook/detr-resnet-50")

        >>> # prepare image for the model
        >>> inputs = image_processor(images=image, return_tensors="pt")

        >>> # forward pass
        >>> outputs = model(**inputs)

        >>> # the last hidden states are the final query embeddings of the Transformer decoder
        >>> # these are of shape (batch_size, num_queries, hidden_size)
        >>> last_hidden_states = outputs.last_hidden_state
        >>> list(last_hidden_states.shape)
        [1, 100, 256]
        ```NrŪ