search

openai / CLIP

CLIP (Contrastive Language-Image Pretraining), Predict the most relevant text snippet given an image
MIT License
25.68k stars 3.29k forks source link

YFCC100M Caption Preprocessing #103

Open normster opened 3 years ago

normster commented 3 years ago

Hi,

I'm trying to reproduce the YFCC100M results and would like to know how image captions were preprocessed during training. For instance, how was the caption for the following sample extracted?

Title: denise%27s+peanut+chicken Description: recipe+here%3A+%3Ca+href%3D%22http%3A%2F%2Fallrecipes.com%2FRecipe%2FDenises-Peanut-Chicken%2FDetail.aspx%22%3Eallrecipes.com%2FRecipe%2FDenises-Peanut-Chicken%2FDetail.aspx%3C%2Fa%3E%0Ai+added+1%2F2+teaspoon+of+sriracha+hot+chili+sauce+and+used+1+TBS+of+chunky+peanut+butter+in+place+of+the+2+cups+of+peanuts.

Best, Norman

naveenkumarmarri commented 3 years ago

You can decode as follows.

import urllib.parse
title = "denise%27s+peanut+chicken"
print(urllib.parse.unquote_plus(title))
# prints the following
"denise's peanut chicken"
TonyLianLong commented 3 years ago

@jongwook In addition to the exact parsing method, could you tell us how a title is merged with the corresponding description text?