search

robodhruv / visualnav-transformer

Official code and checkpoint release for mobile robot foundation models: GNM, ViNT, and NoMaD.
http://general-navigation-models.github.io
MIT License
610 stars 78 forks source link

Did you undistort the images in Go-Stanford dataset before training ViNT model ? #26

Closed han-kyung-min closed 4 months ago

han-kyung-min commented 6 months ago

Hello,

I have a question about your training/testing dataset. It seems that the training dataset contains both rectified (SCAND) and distorted images (Go-Stanford). Did you undistort the images in the Go-Stanford dataset before training your models to standardize the image rectification?

Thanks

ajaysridhar0 commented 4 months ago

We center-cropped the images from the Go-Stanford dataset to have the same aspect ratio as the other images in the combined dataset. By including rectified and distorted images in the training set, we expect our policy to generalize to more camera types.