[dahart]

Which NNs and what tasks are you thinking of? DLSS, for example, reconstructs texture, along with other denoisers, as well as all video models. Are you thinking only of object recognition?

Why do you believe training on dithered images would learn how to see through any other kinds of image modification? In my experience, image NNs learn only what you explicitly show them, and they don’t learn concepts they’re not trained on. All the denoisers I’ve ever seen have trouble handling subtle changes in pixel filtering they haven’t seen in the training data, and they don’t easily learn how to denoise things like clouds or reflections. I’d be willing to bet that if you trained on halftoned images, NNs would not be able to handle Floyd Steinberg dither. They just don’t generalize automatically.

[ipunchghosts]

Eero simmoncellis group showed that specific network modifications allow the networks to extrapolate beyond noise they were trained in.

NNs can certainly learn beyond what they were trained for as new architectures have demonstrated. See benchmarks for any if the dan hendryks imagenet corruption dataset benchmarks (imagenet c r and a). Data + architecture work together to enable.

I have work using neural networks trained on synthetic aperture sonar SAS imagery using language only. The SAS datsets are extremely hard to come by. The dataset I used was from parts if world I can assure you we're never in the training set. My approach exceeds what clip models can do zero shot wise. Therefore, the network classified imagery if a modality and environment 100 percent not guaranteed to be in the training data.