I have raised this in another topic. But finding clones and copyright infringements based on the images seems almost impossible to me. What I think should be done (and you may need some bright minds for that, but I believe that it is possible).
Johnny_Bit has raised a couple times that you can compare meshes with cloudcompare:
It is a program that takes two meshes and compares them to each other and sees how much of the models ‘overlap’. This is a clear indicator that it is either a remix (part of the model overlaps within say 99%) of a blatant reupload (complete model overlaps within that margin).
But also look at face recognition, this is also some kind of cloud-compare software where the overall geometry and relative distances are compressed down into a digital fingerprint of somebody’s face. If you have a camera shot or even a photo of that same person, you can create a new fingerprint and if it matches within a certain margin with another fingerprint it is almost certainly the same person.
Even if that person remixes themselves (say, get a nosejob to evade the police) face recognition still works based on the other remaining features.
So what I propose needs to be done is create this fingerprint for all uploads (preferably open-source so other model sharing websites can create them as well) and easily compare it to the database to find duplicates and flag them, or find remixes and flag them if they are uploaded as original works.
I raised this in another topic, but if you search ‘octopus’ and set to ‘exclusive’ (and therefore ‘original works’), you get these results:
And it goes on and on. All of them are remixes of the same model and all of them are posted as original and even exclusive. Something has to be done imo