I can only see vague parts of a whole, like my object-detector neural net layers are working, and I can visualize lower layers of the network, but not the top l

I can only see vague parts of a whole, like my object-detector neural net layers are working, and I can visualize lower layers of the network, but not the top layer. Like I can't remember my own children's whole face, but I'd instantly recognize it if I see it. It makes me wonder if people's 'visual memory' isn't some for of GAN, where people with good visual memory just have great generative networks with synthesize details which aren't there based on broad parameters. Because when I'm lucid dreaming, or partly asleep, I can have these vivid experiences where I can see imagined objects with perfect clarity, stable detail, even color, something I'm not able to do when normally conscious. Like if I think of red lips, or a red apple right now, I don't really perceive the 'red' -- its abstract. Perhaps the real issue is a 'visual cache' that expires too quickly. Like if your brain is using GAN-like techniques to synthesize imagery from learned network weights, maybe some people have a better working cache, and so it disappears less quickly -- is more persistent. I tend to think abstractly and my mind often gets filled with a bazillion ideas when I'm trying to imagine things, so perhaps I'm experiencing a kind of 'cache thrashing' that I don't get when sedated.