Programs for distinguishing photos from computer graphics

[damienaleksandr]

Hello! This is my first post, so immediately I want to express my gratitude to the site and note that the site is very necessary because it introduces the creativity of a huge number of people involved in architectural graphics!

Computer graphics can often not be distinguished from genuine photography, so the question is: are there programs that can accurately determine: an image - a clean picture, 2D or 3D graphics or a photo, videos created with modification in Photoshop or other editors? Of course you also want to know how to accurately distinguish one render from another, for example in V-Ray it is made, in the Corona or in Lumion.

Necessity in this case, for example, when you need reliable information about the subject.

I use an online translator, so I apologize if there are errors in my text.

[notamondayfan]

Do you mean to find actually photography as reference images?

 

I don't know of any software personally, but if you find any please share!

 

Best way is to find work with credits. You can then check their credentials as to whether they're a photographer or CGI artist.

[damienaleksandr]

Hello Dean Punchard, thank you for participating! It is possible that there are no such programs (there are a lot of reasons). But I think that in the computer field there are specialists who can objectively distinguish a photo from 3d or a real film from a film made on a computer. On the Internet, probably, you can find the answer, but not yet.

Sorry for my english, I use an online translator.

[Corey Beaulieu]

Damien,

 

What is the complete goal, here? I think that while you pose an interesting question, the final answer will be your own eyes. Software, while interested in an answer for fun, may not yet be interested at the commercial level.

 

Personally, I feel that it is "easy" to tell the difference. And I say that knowing that there is unbelievable amounts of incredible work out in the public. But, rendering, especially at the architectural level, is overly clean, overly glowy, and overly expensively designed for real life. Using your eye, you should look for too much detail in certain areas while too little detail in other more practical areas.

 

I do, again personally, feel that there is a certain obviousness to the various render engines and software out there, but it isn't a perfect science. Mostly, if you look at a lot of architectural photos, you will intuit the differences. As a digital artist myself, I see the modeled furniture that we all want to buy and use and I see the over-use of Vray-dirt and the exceptionally bright GI of Corona. I also see an oddity to the scale of Maya and C4D, but a lot of this is impossible to describe.

 

Your own gut instinct may be the software you are looking for if you are only serving a curiosity. If you have a larger purpose, then I am not certain there is a software available.

 

I hope this answers your question to some degree.

 

Corey

[damienaleksandr]

Сorey Beaulieu, thank you! No, I can not answer my question because I do not know the essence of the computer world. Of course, it is not always possible to easily prove a photo or render, especially if the author wanted to issue one for the other. In VIP on one site there was a funny case, the reference with the photo of the car GAZ 12 began to be criticized, taking the photo for rendering - and the light is bad and the materials are bad and in general the model is "curve". To illustrate the topic I put photos of cars in the forest with snow, on the same site opinions were 50% of the photo, 50% render, and very good 3d artists rated it. For these examples, we can assume that owning an objective method of distinguishing a photo / video from a render can be a profitable business.

How to attach pictures - tell me, please - I can not

[damienaleksandr]

I tried to find an answer to my topic on the Internet and was surprised to find that this topic interests not only me. There are sites on which it is proposed to test it by tests. Interestingly, of the 10 submitted photos and renderers, 10 responses are incorrect. The reason for the incorrect answers is that the renderers were done there with properties characteristic of the photo and vice versa. These sites were devoted to computer graphics. Naturally, in the tests took part 2 and 3d artists. This means that the visual method does not guarantee the truth. On the Internet, I saw Jeffrey's Exif Viewer, he defines the camera for free with what camera it was made, distance, date and exact geographic address. Naturally, when this data is not in the properties of the file. In the hope that such a specialist will tell something on my question, I turned to him. But, unfortunately, he replied that his area is working with a photograph, and this question is not part of his field of activity.

[damienaleksandr]

Is it possible to determine the render with the proofs in which the picture is made? (Virey, Lumion, Mental Ray and the program - 3d Max, AutoCad, Lumion and others)

And the second question - on what technical, objective parameters determine the quality of the render (visualization)? But if the quality issue has already been considered at the forum, then please give a link to it.

[James Vella]

Your best bet is metadata - without that no, you have no information to work out the source of the image.

 

For example a Camera .CR2 file has metadata that tells you what camera took the photograph, camera details, date, even who took the image if they have registered the camera under their name etc.

 

Rendered images usually dont contain this type of metadata, and gets more complicated as it goes through the post production phase as many other .jpg/.tif/software/filters/custom processes etc have been used to manipulate & formulate that final image.

 

Your question would be akin to say asking what type of stencil was used to create a painting made by a traditional artist - did he build a plastic stencil, a paper cutout, a piece of cardboard? Without even knowing the name of that artist you could not even start to draw comparisons as to what type of workflow was used, maybe that artist does not even use stencils and has a very steady hand... you can start to see the dilemma here.

 

On a technical stand point only observation will get you so far such as: I recognize this material set, this lighting style, this interior design, or patterns of workflow (such as artist X always uses this type of workflow so I would guess its probably artist X creating this, but if artist Y uses their work as inspiration and does a very good job at emulating it you would not know without some ownership). Maybe 2 years ago I would say this looks like a Keyshot rendering, or a Vray rendering, or Corona rendering because of the lighting, or some technical limitation which is common to that render engine - but these days the quality gap is closing and all photo quality renders are producing amazing results. Only 1 currently stands out to me at the moment which is Fstorm, some of the renders I see from that community do look quite like photos at times - but this is besides the question in the OP, just my bias from my personal experience. Without that experience I doubt you will find any other way of finding the true source of the imagery without the help of some AI built specifically for this, and even then wont be correct a lot of the time based on what I mentioned earlier regarding the post production phase.

Edited August 4, 2018 by redvella