Fig. 5From: Lifelogging caption generation via fourth-person vision in a human–robot symbiotic environmentDetails of the attention module and the decoder module of UpDpwn model [20]Back to article page