A modified U-Net with HED-inspired decoder side layers and dilation convolution extracts skeletons from object shape pixels and scores 0.77 F1 on competition test data.
An intriguing failing of convolutional neural networks and the coordconv solution
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
-
SkeletonNet: Shape Pixel to Skeleton Pixel
A modified U-Net with HED-inspired decoder side layers and dilation convolution extracts skeletons from object shape pixels and scores 0.77 F1 on competition test data.