Hello, may I ask you some questions? I do medical image analysis about segment-correction-net HOT 4 OPEN

So ur final network input is an image and some bbox for it?

When u train a normal segmentation network does it work wekk?

Notice that: the no-correction module is almost the same as Deeplabv3+. U only add bbox encoder to the network

from segment-correction-net.

So ur final network input is an image and some bbox for it?

When u train a normal segmentation network does it work wekk?

Notice that: the no-correction module is almost the same as Deeplabv3+. U only add bbox encoder to the network

As pointed out in your paper, the incoming network includes medical images and bbox, as well as a small part of complete boundary annotation. But I don't know if your method will work well, because the target and background in the image are very similar. I don't know if your semi supervised method or full supervised method would be better. Thank you for your reply

from segment-correction-net.

I see the challenge. I think our method fits more with natural scenes.

from segment-correction-net.

I see the challenge. I think our method fits more with natural scenes.

OK, thank you for your reply. Also, did you know that someone copied the complete code from your paper? If so, can you provide the relevant website? thank you!

from segment-correction-net.