MT-CP improves multi-task dense prediction with cross-task feature coherence and a dynamic loss-prioritization scheme, reporting new state-of-the-art results on NYUD-v2 and PASCAL-Context.
Gradnorm: Gradient normalization for adaptive loss balancing in deep multitask networks
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
citation-role summary
background 1
citation-polarity summary
fields
cs.CV 1years
2024 1verdicts
CONDITIONAL 1roles
background 1polarities
support 1representative citing papers
citing papers explorer
-
Optimizing Dense Visual Predictions Through Multi-Task Coherence and Prioritization
MT-CP improves multi-task dense prediction with cross-task feature coherence and a dynamic loss-prioritization scheme, reporting new state-of-the-art results on NYUD-v2 and PASCAL-Context.