Hi, I am wanting to fine tune a pretrained image segmentation network on a new dataset. Specifically, I want to finetune the DeepLabV3 with a resnet101 backbone pretrained on the ADE 20k data set. Following the demo_deeplab.ipynb on gluon-cv, I download the model: model = gluoncv.model_zoo.get_mo…

Change Conv2D layer on pretrained network for Image Segmentation

lsps October 28, 2018, 2:41pm 16

Okay great. Will check out that github issue.

Thanks for the help!

1 Like