Dank Learning: Generating Memes Using Deep Neural Networks

This is non the futurity I was promised.

From Stanford University's Computer Science mavens via arXiv.org:
Abel L Peirson V, E Meltem Tolunay

(Submitted on viii Jun 2018)

We innovate a new meme generation system, which given whatever icon tin create a humorous too relevant caption. Furthermore, the organisation tin survive conditioned on non exclusively an icon only likewise a user-defined label relating to the meme template, giving a grip to the user on meme content. The organisation uses a pretrained Inception-v3 network to provide an icon embedding which is passed to an attention-based deep-layer LSTM model producing the caption - inspired past times the widely recognised Show too Tell Model. We implement a modified beam search to encourage form inwards the captions. We evaluate the character of our model using perplexity too human assessment on both the character of memes generated too whether they tin survive differentiated from existent ones. Our model produces master copy memes that cannot on the whole survive differentiated from existent ones.

Comments:	Stanford CS 224n Project
Subjects:	Computation too Language (cs.CL); Learning (cs.LG)
Cite as:	arXiv:1806.04510 [cs.CL]
	(or arXiv:1806.04510v1 [cs.CL] for this version)

arXiv download page (9 page PDF)