Skip to content
王佳欣 edited this page Jun 27, 2018 · 3 revisions

Welcome to the torchseg wiki!

pspnet + edge

dilate semantic edge to improve final results

pspnet + global

pool/dilated convolution on raw semantic segmentation to form a final semantic segmentation the final semantic segmentation can be class/category segmentation

pspnet + dict learning

combine pspnet with dict learning

pspnet + random number filters

pspnet with [1,2,4,...]*class_number filters, some what like fractal network

Clone this wiki locally