Sequential encoder-decoder distillation provides robust initialization for lightweight image compression autoencoders, followed by standard training that preserves reconstruction quality better than direct optimization in early epochs.
For both architectures, gS s (·) = gT s (·) and EB S (·) = EB T (·), and in the Hyperprior case also hS s (·) = hT s (·)
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Efficient training for compact compression models via sequential distillation
Sequential encoder-decoder distillation provides robust initialization for lightweight image compression autoencoders, followed by standard training that preserves reconstruction quality better than direct optimization in early epochs.