The way the application works on the surface is quite straightforward. Then the algorithm looks for cranes and detects every single pixel of the object. Finally, another algorithm – similar to the ones used in photo editors like Photoshop – replaces those pixels with new ones. Eventually the user gets a beautiful picture without unwanted cranes.
@ishoottheworld@hello_roman@abadesi haha, I like the idea :) Even if it's slightly more involved than what I had in mind – removing clouds would be easy, but you also have to remove the gloom :) Objects we were considering to remove next are people and pigeons.
It's crazy that this is a real need. But since I live in a big city with cranes everywhere I thought I would try it. I tried this image , but it failed to find a crane. I guess the image quality is not good enough but I thought the crane was pretty visible.
To be fair I tried using other images where the crane was up close and it did a pretty good job at removing it and matching the background.
@tostartafire I am afraid your crane is too small for the network to see it (especially that we have to downscale the image before we passing it to the network).
@tostartafire In terms of problems with small cranes I recommend to cut the interesting part of the picture (the part which contains cranes) and pass just it to the prettycitty app engine - please check result on the cropped image: https://imgur.com/FnKtuAl
Tried a few times... didn't do anything... after the fun "loading" messages... it just goes back to "upload picture"... will try again when it works though!
Hustle Crew Academy
PrettyCity
Hustle Crew Academy
PrettyCity
PrettyCity
Symphony
PrettyCity