Exploring knowledge fusion of multi-modal embeddings consisting text, image and WordNet knowledge-graph concepts.
We have provided the source code (in tar.gz) available at: https://drive.google.com/file/d/1oTI0LAnGrGO0upCM1f0ZjUHrwlG3zD3t/view?usp=sharing
To extract the source code files and dataset, just run tar -xvzf rojak.tar.gz. You should now have the extracted folder: Rojak, and you can view the codebase on ProjectROJAK.ipynb.
For convenience, we have also hosted a comprehensive notebook at https://kenghweeng.github.io/Rojak/