Code used in the paper "Learning to Learn from Web Data through Deep Semantic Embeddings" ECCV 2018 MULA Workshop
A blog post explaining the work is available here.
-
The code in the folders named with a text embedding name is to train text models and compute the text embeddings.
-
The code in googlenet_regression folder trains a CNN (PyCaffe) to regress those text embeddings from images.
Read the paper or the blog post for more details.