OrganSegC2F_v1.0

 - Original release.

OrganSegC2F_v1.1

 - OrganSegC2F/DataF.py:

     . Added the 3-slice-predict-3-slice strategy, which improves the performance.

     . Changed the distribution of the random margins added to four sides.

 - OrganSegC2F/coarse_fusion.py:

     . Changed the definition of thresholds (this does not impact the results).

 - OrganSegC2F/oracle_fusion.py:

     . Added codes for 3-slice-predict-3-slice.

     . Changed the definition of thresholds (this does not impact the results).

 - OrganSegC2F/coarse2fine_testing.py:

     . Added codes for 3-slice-predict-3-slice.

     . Changed the definition of thresholds (this slightly improves the results).

 - OrganSegC2F/prototxts/

     . Added the prototxt and deploy files for 3-slice-predict-3-slice.

     . Renamed the files (the corresponding changes were also made in other codes).

OrganSegC2F_v1.2

 - OrganSegC2F/DataC.py:

     . Added the 3-slice-predict-3-slice strategy, which improves the performance.

 - OrganSegC2F/run.sh:

     . Changed the parameter list for some modules.

     . Slightly changed the naming method for snapshots and testing results.

 - OrganSegC2F/coarse_training.py

     . Allowed 3-slice-predict-3-slice coarse-scaled training.

 - OrganSegC2F/coarse_testing.py

     . Allowed 3-slice-predict-3-slice coarse-scaled testing.

 - OrganSegC2F/prototxts/

     . Added prototxt and deploy files for 3-slice-predict-3-slice training and testing.

OrganSegC2F_v1.3

 - OrganSegC2F/run.sh:

     . Changed the default TRAINING_MARGIN to 20 for better results.

 - OrganSegC2F/coarse2fine_testing.py

     . Fixed a small bug involving the path definition (no impacts on the results).

     . Fixed a small bug which will exit the program when all tasks are finished beforehand.

OrganSegC2F_v1.4

 - OrganSegC2F/utils.py

     . Fixed a bug in data split (fortunately, the split on the NIH dataset is correct).

     . Allowed a flexible definition of between organ_ID and label.

OrganSegC2F_v1.5

 - OrganSegC2F/utils.py

     . Fixed a bug in data split (the split on the NIH dataset is correct).

OrganSegC2F_v1.6

 - OrganSegC2F/utils.py

     . Fixed a bug in finding most recent snapshots and results.

 - OrganSegC2F/oracle_testing.py

     . Fixed a bug in the logical order of codes (no impacts on the current results).

OrganSegC2F_v1.7

 - OrganSegC2F/run.sh

     . "Lazy Mode" added, which allows to run everything within 1 click (see README for details).

     . "PLANE" variables can be set as "A": training/testing all three planes orderly.

 - OrganSegC2F/init.py

     . Sliced image data are not truncated any more (no impacts on the current results).

OrganSegC2F_v1.8

 - OrganSegC2F/run.sh

     . Fixed a small bug in "Lazy Mode" (log files not automatically moved to snapshot folders).

 - OrganSegC2F/prototxts/

     . Slightly modified network structures:

       * Average accuracy on the NIH dataset improved by 0.3% - 0.5%;

       * Computational overhead (time and memory) increased by less than 2%.

     . Original prototxts files provided in OrganSegC2F/prototxts_original/ .

OrganSegC2F_v1.9

 - OrganSegC2F/coarse2fine_testing.py

     . Slightly changed the way of taking slices according to the previous round.

       * To be consistent to the oracle_testing.py;

       * Little impact on the overall average accuracy (difference < 0.01%).

OrganSegC2F_v1.10

 - OrganSegC2F/coarse2fine_testing.py

     . Fixed a small bug, which may appear when the output volume is empty.

       * No impact on the overall average accuracy on NIH.

OrganSegC2F_v1.11

 - OrganSegC2F/run.sh

     . Added an export to $LD_LIBRARY_PATH, decreasing the probability of import errors.

 - OrganSegC2F/coarse_training.py

 - OrganSegC2F/fine_training.py

     . Fixed a hidden trouble which may cause file writing problems by accident.

 - OrganSegC2F/prototxts/

     . Undid the modifications in v1.8, as we observed unstable performance in other organs.