boring script out dir

config.yml

@@ -13,4 +13,5 @@ felix_labels_dir_2: "1Felix and Rich make models/Training dataset Tiffs/Training
 wahab_labels_dir: "1Felix and Rich make models/Training dataset Tiffs/Training set 1"
 
 # Where random stuff from the scripts should go, relative to the root of the repository
-script_output: "script_output/"
+script_output: "script_output/"
+boring_script_output: "script_output/boring_stuff/"

fishjaw/util/files.py

@@ -163,3 +163,17 @@ def script_out_dir() -> pathlib.Path:
     if not retval.is_dir():
         retval.mkdir()
     return retval
+
+
+def boring_script_out_dir() -> pathlib.Path:
+    """
+    Get the directory where the output of the extra scripts is stored,
+    creating it if it doesn't exist
+
+    :returns: Path to the directory
+
+    """
+    retval = pathlib.Path(__file__).parents[2] / "boring_script_output"
+    if not retval.is_dir():
+        retval.mkdir()
+    return retval

scripts/README.md

@@ -20,21 +20,3 @@ Roughly in the order that you might run them:
 - `arch_summary.py`: summarise the architecture of the model
 - `inference_example.py`: run inference on some data
 - `mesh.py`: create and visualise a mesh from the inference
-
-Some others that you probably won't run at all, because they're basically not interesting
-- `check_loss.py`: check that the loss function behaves as you might expect,
-                 including how it behaves at chance and perfect performance.
-                 So that the expected loss is 0.0 at perfect performance and
-                 $\frac{1}{1+\frac{1}{2}\alpha + \frac{1}{2}\beta}$ at chance performance,
-                 This script doesn't use a sigmoid or softmax activation fcn.
-                 Creates a plot named `chance_loss.png`.
-- `find_rear_jaw_centres.py`: find the ZXY locations of the jaw in of the rear jaw dataset. We need this
-                              because we want to crop our jaw from the last labelled slice backwards (towards
-                              the tail) so that our cropped out regions of interest don't contain any unlabelled
-                              jaw voxels. This script makes some decisions and crops the jaws.
-- `plot_cropped.py`: check that the cropping code gives us the correct regions.
-                   This is intended to be a check that the cropping code correctly
-                   gives us the correct slices for the rear jaw DICOMs, since it's
-                   important that there's no unlabelled jaw in the training data.
-- `lines_of_code.py`: plot the total lines of code in this respository over time. Just for fun!!!!
-- `find_occupancy.py`: find how much of each DICOM is occupied by jaw voxels, for both the cropped and uncropped DICOMs.