Open GYengera opened 5 years ago
Hi, @GYengera
If you give a look under the live demo http://nvidia-research-mingyuliu.com/gaugan You can find the link to the implementation used: https://github.com/kazuto1011/deeplab-pytorch.
From that page, it seems like the network was trained also on COCO-Stuff which has some outdoor segmentation labels
Hello,
I had a question regarding which pretrained DeepLab model was used for computing segmentation maps for the Flickr landscape images. The DeepLab model trained on the Pascal VOC dataset appears to label most of the landscape images as background. So was a model that was pretrained on another dataset utilized or did you train a new semantic segmentation model on landscape images?