Ask HN: ML-Experts) How Did Transformers Improve Your Computer Vision Task

Hi all, I try to keep up to date with all the ML/ANN tech and of course Transformers/Attention as we know had a HUGE (from what I can tell) impact on language and (other seq2seq task ?)

I see such great use/papers like FXNET using Fourier transform and other MLP mixer techniques to improve vision task. It would be super interesting to hear from say "Tesla" or other leading self-driving tech companies, how much have Transformers/Attention improve your stack/solution ?

from a layman