Coming to the main model, image captioning architecture consists of three models:

A CNN: used to extract the image features A TransformerEncoder: The extracted image features are then passed to a Transformer based encoder that generates a new representation of the inputs A TransformerDecoder: This model takes the encoder output and the text data (sequences) as inputs and tries to learn to generate the caption.

CNN extract features >> Tranformer encoder (new representation of CNN output) >> TransformerDecoder takes (transformer encoder outputs + text data (in integer sequence format) and learns to generate captions corresponding to imgs)