Hello, thanks for your great work!
I’ve been testing EVF-SAM2, and I can easily reproduce performance reported on paper. But when I visualize the output masks, I noticed that they often look less smooth and more jagged compared to those from the original SAM2.
Have you experienced a similar issue? If so, could you share what you think might be causing it? (e.g., the ratio between CE loss and Dice loss)
Hello, thanks for your great work!
I’ve been testing EVF-SAM2, and I can easily reproduce performance reported on paper. But when I visualize the output masks, I noticed that they often look less smooth and more jagged compared to those from the original SAM2.
Have you experienced a similar issue? If so, could you share what you think might be causing it? (e.g., the ratio between CE loss and Dice loss)