Part 6 - PIX2PIX Panos (Done on 13th June, 2017)

PIX2PIX can also recognize gray scale images, so next, we got the data from Google street view and its depth maps, and trained the model with following image pairs:

alt text alt text alt text

Delft_Panos_512_Unfolded, 450 images, dataset here: 18_Delft_Panos_512_Unfolded.rar

One change is that, we set the size of images to 1024*512 and adjusted the script to get larger outputs. Then we used simple digital models to simulate the outcome. Video Here:

alt text alt text alt text

Albaicin_Panos_512_Unfolded, 465 images, dataset here: 19_Albaicin_Panos_512_Unfolded.rar

Video Here:

alt text alt text alt text

Mexico_Panos_512_Unfolded, 285 images, dataset here: 20_Mexico_Panos_512_Unfolded.rar

Video Here:

alt text alt text alt text

Death Valley_Panos_512_Unfolded, 720 images, dataset here: 21_Death Valley_Panos_512_Unfolded.rar

Video Here:

Draw Your Own Street

Script to get depth maps