1 paper · 1 filter
Esin Koyuncu, Timofey Solovyev, Johannes Sauer +2
Learned image compression has a problem of non-bit-exact reconstruction due to different calculations of floating point arithmetic on different devices. This paper shows a method t…