Thank you for your outstanding work in polyp segmentation!
Following your code and settings, I successfully reproduced the remarkable results from your paper. Below are the test results on the ETIS-LaribPolypDB dataset (85.87), and the mDice, which closely matches the results in the paper (86.51).
However, I have a question. The mDice you calculated (85.87) is the average of the background Dice (98.37) and the foreground Dice (73.37). This differs from mainstream methods like PraNet, where people prefer to use the so-called foreground Dice (73.37) as the benchmark for comparison, rather than the average of foreground and background dice.
What are your thoughts on this issue? Thanks again.
Thank you for your outstanding work in polyp segmentation! Following your code and settings, I successfully reproduced the remarkable results from your paper. Below are the test results on the ETIS-LaribPolypDB dataset (85.87), and the mDice, which closely matches the results in the paper (86.51).
However, I have a question. The mDice you calculated (85.87) is the average of the background Dice (98.37) and the foreground Dice (73.37). This differs from mainstream methods like PraNet, where people prefer to use the so-called foreground Dice (73.37) as the benchmark for comparison, rather than the average of foreground and background dice. What are your thoughts on this issue? Thanks again.