Not defending OP who clearly has a screw loose. Also it is too early to know what ROOST actually plans on doing, no details have been published.
I just wanted to note that you couldn't train a model on the cryptographic hashes of images because the purpose of that method is to map the input to 1/3
I just wanted to note that you couldn't train a model on the cryptographic hashes of images because the purpose of that method is to map the input to 1/3
Comments
Also it is sort of possible to generate images from a deep image classifier, but they are very poor. Certainly, the original dataset cannot be recovered. 3/3