>> Нейронки - это тупик!
> Применение "машинного обучения" в текущем виде - это тупик. А нейронки как
> раз таки очень эффективны для обработки изображений (это доказано).

По изображениям есть такое мнение

https://arxiv.org/abs/1805.04025
At present Deep Nets do very well on specific types of visual tasks and on specific benchmarked datasets. But Deep Nets are much less general purpose, flexible, and adaptive than the human visual system. Moreover, methods like Deep Nets may run into fundamental difficulties when faced with the enormous complexity of natural images which can lead to a combinatorial explosion.

что сделать что-то похожее на человеческое зрение на глубоких сетях не получится (и привет Маску).

на практике другие подходы частото работают лучше

https://arxiv.org/pdf/1805.05837
Comparing LBP, HOG and Deep Features for Classification of Histopathology Images

SVM achieves the highest accuracy of 90.52% using local binary patterns as features which surpasses the accuracy obtained by deep features, namely 81.14%.

но это не модно-молодежно и бабки на это не получишь

https://thegradient.pub/the-limitations-of-visual-deep-learn.../
Now it is difficult to publish anything that is not neural network related. This is not a good development. We suspect that the field would progress faster if researchers pursued a diversity of approaches and techniques instead of chasing the current vogue. It is doubly worrying that student courses in AI often completely ignore the older techniques in favor of the current trends..