Dank Learning: Generating Memes Using Deep Neural Networks
This addresses the problem of automated meme creation for social media users and content creators, though it is incremental in building on existing captioning models.
The researchers tackled the problem of automatically generating humorous and relevant captions for memes from any image, with the ability to condition on user-defined labels for content control. Their system produced original memes that humans could not reliably distinguish from real ones.
We introduce a novel meme generation system, which given any image can produce a humorous and relevant caption. Furthermore, the system can be conditioned on not only an image but also a user-defined label relating to the meme template, giving a handle to the user on meme content. The system uses a pretrained Inception-v3 network to return an image embedding which is passed to an attention-based deep-layer LSTM model producing the caption - inspired by the widely recognised Show and Tell Model. We implement a modified beam search to encourage diversity in the captions. We evaluate the quality of our model using perplexity and human assessment on both the quality of memes generated and whether they can be differentiated from real ones. Our model produces original memes that cannot on the whole be differentiated from real ones.