Hacker News Reader

Comment by skzv | Hacker News Reader

I can understand the concern. It can mislead people about what the past really looked like.

An example: this same model colorized the Golden Gate white: https://camo.githubusercontent.com/3487a03626fd86d4f8c814636...

In my case I wasn't too concerned, since I was sharing the photos with my family, who understood they were artificially colorized. My father even asked me if it was possible to choose specific colours in articles of clothing, since it got those wrong. Nonetheless, to me and my family, it made our old photographs seem so close and alive, which made it worth it.

A good compromise may be to mark these retouched photos with a small watermark indicating that they have been altered. By the way, I think the colab instances running this model have the option to add watermarks for this reason (https://colab.research.google.com/github/jantic/DeOldify/blo...).

View on HN

> My father even asked me if it was possible to choose specific colours in articles of clothing, since it got those wrong.

Did you figure out a way to achieve this other than repeat attempts or recoloring in post?

Is there a setup that retrains with user input such as by inputting an image with an object manually colored the model will then recognize it in later input and color similarly?

It seems like this could also help with persistence of colors in videos, training with the previously completed frame step before producing the next.

Silly question, but was the bridge definitely painted red at that stage of its life?

Panchromatic vs orthochromatic was still a choice even in the 70s. Spotmatics shipped with an ISO marker dial that included color, ortho, and panchro, 30 years after the Golden Gate Bridge was built.

Ortho bw — cheap 1930s snapshot — would show a red bridge as white, but if it was fancy panchro then I’d be more inclined to agree with the AI, given the luminosity.

There was this incident from last month: https://www.aljazeera.com/news/2021/4/11/cambodia-condemns-v...

Cambodia has called on US media group Vice to withdraw an article that featured newly-colourised photographs of victims of the Khmer Rouge, saying the images are an insult to the dead because some had been altered to add smiles.

“We urge researchers, artists and the public not to manipulate any historical source to respect the victims,” [Cambodia’s Ministry of Culture] said.

I don't think it was just the colorizing that was the problem. He was changing their expressions into smiles!

I wonder at what point the photo would be colorized correctly. The golden gate bridge is almost certainly in the training data if it's random photos. Maybe it knows to color completed things?

what about photos that were taken in colors but then get edited to look a certain way (most of photos on the internet)?

and what about those space photos?

Even every photo you take with a smartphone these days is altered by a machine learned model, put through some filter, and tuned in some way, even if it's subtle.

I think that there’s a general understanding that cameras don’t match 100% what the eye sees. That’s different from a white golden gate.

people see a white golden gate and they immediately know that's not a realistic representation. They don't look at photos heavily touched or space photos with that expectation.

Was the Golden Gate bridge painted at that point? [1] says "some" of the steel came with red primer.

https://www.npr.org/2011/04/26/135150942/the-golden-gate-bri...

Reply on news.ycombinator.com