The process of image caption

[cm-xcju]

Is there the process code about image caption? I find a "captions_okvqa.json" file in the assets about image caption, but I do not find the process code about this file

[MIL-VLG]

The caption is extracted using the VinVL caption model, which is identical to a previous repo PICa.
Alternatively, you can use the code in VinVL to extract caption for each image by yourself