search

cshizhe / VLN-DUET

Official implementation of Think Global, Act Local: Dual-scale GraphTransformer for Vision-and-Language Navigation (CVPR'22 Oral).
114 stars 7 forks source link

ViT feature extraction code #10

Closed heyzude closed 2 years ago

heyzude commented 2 years ago

Hi, and thanks for your great work and for sharing it with us :)

I wonder if you could also share the code used for extracting image features from ViT.

I also wonder what folder you extracted features from. There are undistorted_color_images, matterport_color_images, and matterport_skybox_images folders for each scan.

Thanks!

cshizhe commented 2 years ago

Hi, the images are rendered with the Matterport3DSimulator which uses the matterport_skybox_images folder. Check this for feature extraction codes.