Dank Learning

Crypt | 2020
personMembers
descriptionDescription

Our model consists of an encoder-decoder scheme. The encoder is a deep convolutional neural network (CNN) that that takes images as inputs and produces fixed-length vector embeddings. To capture the image embeddings, we use a CNN model Inception-v3 pretrained on the ILSVRC-2012-CLS image classification dataset. We explore different encoding schemes and make comparisons based on that The decoder consists of a transformer model with multi-headed semantic attention that generates the captions We introduce and test several variants of the Transformer model and evaluate their outputs. Lastly we'll have a Web App using Django

local_offerTags
Deep learning
Memes
CNN
Webdev