Question about how to invoke scene graph in caption_model

[xixiareone]

hi! In caption_model, I don't see how you invoke the code of the scene graph for each video you generated before, but instead use the data in sum data. Where did you invoke the scene graph you generated? thank you very much!