missing code in the paper

[YU-deep]

I cannot find the corresponding codes about "Memory Bank Update with Confidence and Dissimilarity", please tell me where is are they? Thank you so much!

[jguinea]

I have the same issue. I want to use the memory bank with 3D volumes, processing all orientations as explained in the paper.

The memory bank and its similarity scores only seem implemented in func_2d/function.py, which works with the REFUGE dataset (2D) but not with the BTCV dataset (which is the most similar to my data).

In func_3d/function.py, the volume is propagated slice by slice as if it were a video, using SAM2's propagate_in_video function. If I'm not mistaken, it uses a FIFO queue as the memory bank and only processes the volume in one orientation.

Is any script available where the self-sorting memory bank is applied to 3D images in all 6 directions?

Thank you in advance.