FRAME Revisited: An Interpretation View Based on Particle Evolution

FRAME (Filters, Random fields, And Maximum Entropy) is an energy-based descriptive model that synthesizes visual realism by capturing mutual patterns from structural input signals. The maximum likelihood estimation (MLE) is applied by default, yet conventionally causes the unstable training energy that wrecks the generated structures, which remains unexplained... (read more)

Results in Papers With Code
(↓ scroll down to see all results)