Skip to content

Given a concept, select a set of k highly pertinent images.

Notifications You must be signed in to change notification settings

mikeogezi/BabelCLIP

Repository files navigation

BabelCLIP: CMPUT 600 W22 Project

To get a better understanding of the results of this work, you can read my report. BabelCLIP outperforms the previous SOTA by 6.67%.

Prerequisites

  • Install Java (OpenJDK 11.0.14.1; I make extensive use of lambdas and streams)
  • Install Python 3.8.12

Playing with this work

  • Download and configure BabelNet 5.0 indices
  • Download the Java API from the same page as above
  • Download and extract BabelPic-Gold
  • Generate gold_synsets.json:
    • cd java
    • sh run-babelpic.sh
    • cd ..
  • Optionally download and extract the silver splits of BabelPic
  • python pull.py --sample_size 100
  • python run_experiments.py

About

Given a concept, select a set of k highly pertinent images.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages