crossview_localisation icon indicating copy to clipboard operation
crossview_localisation copied to clipboard

how to query the model

Open shokah opened this issue 6 years ago • 2 comments

hi,

first of all, very impressive work!

i'm trying to run the given model on the cvusa data set. i didn't find a simple query method in your code (something like - input:ground image, output: matching sat image), so i tried to make one based on your "validate" section in train.py. but i were unable to recreate the results your got (~70% for the top 1%).

can you upload the script the got your these results?

shokah avatar Aug 20 '18 10:08 shokah

I did write the code to query one ground image. You can write it based on my validation function. Simply change the ground image matrix to one entry. Query one image and a set of images should be the same.

david-husx avatar Sep 18 '18 12:09 david-husx

@shokah hi, have you implemented an image query? If so,can you share the script? thx

DylanMaeng avatar Sep 29 '19 12:09 DylanMaeng