Skip to content

End-to-end Image Captioning Exploits Distributional Similarity in Multimodal Space

Notifications You must be signed in to change notification settings

sheffieldnlp/whatIC

Repository files navigation

End-to-end Image Captioning Exploits Distributional Similarity in Multimodal Space

These are original images for Section 4.2:

  • tsne_initial_boc_4000.png - corresponds to tsne projection of the initial bag of object representation

  • tsne_projected_boc_4000.png - corresponds to tsne projection of the projected bag of object representation

  • tsne_initial_pseudorandom_4000.png - corresponds to tsne projection of the initial pseudorandom representation

  • tsne_projected_pseudorandom_4000.png - corresponds to tsne projection of the projected pseudorandom representation

  • tsne_initial_softmax_4000.png - corresponds to tsne projection of the basic softmax representation

  • tsne_projected_softmax_4000.png - corresponds to tsne projection of the projected softmax representation

  • tsne_initial_pool5_4000.png - corresponds to tsne projection of the original Resnet152 Pool5 representation

  • tsne_projected_pool5_4000.png - corresponds to tsne projection of the projected Resnet152 Pool5 representation

About

End-to-end Image Captioning Exploits Distributional Similarity in Multimodal Space

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published