ContextPrior
ContextPrior copied to clipboard
Some confusion about this work
Many thanks for this great work. After reading your paper. I have some confusion.
- You mention in your paper, you train your model over ADE20k with 480x480 which used to be 512x512 in MMSegmentation. Do you adopt multi-crop testing during the evaluation for this dataset?
- Any plans to release the trained models?