RetinaNet multiscale embeddings improve pattern location accuracy and reduce storage needs versus prior methods on the DocExplore dataset, though they fail on pages with multiple query instances.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2019 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Pattern Spotting in Historical Documents Using Convolutional Models
RetinaNet multiscale embeddings improve pattern location accuracy and reduce storage needs versus prior methods on the DocExplore dataset, though they fail on pages with multiple query instances.