A dual-stage, entropy-guided knowledge amalgamation method trains a compact multi-task student from heterogeneous single- and multi-task teachers without labels, and the student outperforms its teachers in reported experiments.
Student becoming the master: Knowledge amalgamation for joint scene parsing, depth estimation, and more
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2019 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Customizing Student Networks From Heterogeneous Teachers via Adaptive Knowledge Amalgamation
A dual-stage, entropy-guided knowledge amalgamation method trains a compact multi-task student from heterogeneous single- and multi-task teachers without labels, and the student outperforms its teachers in reported experiments.