Miscellanious fixes

README.md

@@ -27,10 +27,6 @@ If you'd like to reproduce this analysis on our data, check out the following do
 - [Preprocessing documentation](./docs/PREPROCESSING.md) for generating pointclouds and SDFs from from our input movies (step 2).
 - [Development documentation](./docs/DEVELOPMENT.md) for guidance working on the code in this repository.
 
-# Using the models
-
-Coming soon
-
 # Quickstart
 
 Use the cytodl api to run any given experiment. For e.g., train a rotation invariant point cloud autoencoder on the PCNA dataset

configs/results/npm1.yaml

@@ -26,4 +26,12 @@ data_paths:
     "/data/npm1/pc.yaml",
   ]
 classification_label: ["STR_connectivity_cc_thresh"]
-regression_label: ["mean_surface_area", "mean_volume", "avg_dists", "std_dists"]
+regression_label:
+  [
+    "mean_surface_area",
+    "std_surface_area",
+    "mean_volume",
+    "std_volume",
+    "avg_dists",
+    "std_dists",
+  ]

configs/results/npm1_64_res.yaml

@@ -3,11 +3,6 @@ image_path: ./morphology_appropriate_representation_learning/preprocessed_data/n
 pc_path: ./morphology_appropriate_representation_learning/preprocessed_data/npm1_64_res/manifest.csv
 model_checkpoints:
   [
-    "./morphology_appropriate_representation_learning/model_checkpoints/npm1/Classical_image_seg.ckpt",
-    "./morphology_appropriate_representation_learning/model_checkpoints/npm1/Rotation_invariant_image_seg.ckpt",
-    "./morphology_appropriate_representation_learning/model_checkpoints/npm1/Classical_image_SDF.ckpt",
-    "./morphology_appropriate_representation_learning/model_checkpoints/npm1/Rotation_invariant_image_SDF.ckpt",
-    "./morphology_appropriate_representation_learning/model_checkpoints/npm1/Rotation_invariant_pointcloud_SDF.ckpt",
     "./morphology_appropriate_representation_learning/model_checkpoints/npm1_64_res/Classical_image_seg.ckpt",
     "./morphology_appropriate_representation_learning/model_checkpoints/npm1_64_res/Rotation_invariant_image_seg.ckpt",
     "./morphology_appropriate_representation_learning/model_checkpoints/npm1_64_res/Classical_image_SDF.ckpt",
@@ -21,24 +16,22 @@ names:
     "Classical_image_SDF",
     "Rotation_invariant_image_SDF",
     "Rotation_invariant_pointcloud_SDF",
-    "Classical_image_seg_64",
-    "Rotation_invariant_image_seg_64",
-    "Classical_image_SDF_64",
-    "Rotation_invariant_image_SDF_64",
-    "Rotation_invariant_pointcloud_SDF_64",
   ]
 data_paths:
   [
-    "/data/npm1/classical_image_seg.yaml",
-    "/data/npm1/so3_image_seg.yaml",
-    "/data/npm1/classical_image_sdf.yaml",
-    "/data/npm1/so3_image_sdf.yaml",
-    "/data/npm1/pc.yaml",
     "/data/npm1_64_res/classical_image_seg.yaml",
     "/data/npm1_64_res/so3_image_seg.yaml",
     "/data/npm1_64_res/classical_image_sdf.yaml",
     "/data/npm1_64_res/so3_image_sdf.yaml",
     "/data/npm1_64_res/pc.yaml",
   ]
 classification_label: ["STR_connectivity_cc_thresh"]
-regression_label: ["mean_surface_area", "mean_volume", "avg_dists", "std_dists"]
+regression_label:
+  [
+    "mean_surface_area",
+    "std_surface_area",
+    "mean_volume",
+    "std_volume",
+    "avg_dists",
+    "std_dists",
+  ]

configs/results/other_polymorphic.yaml

@@ -3,27 +3,27 @@ image_path: ./morphology_appropriate_representation_learning/preprocessed_data/o
 pc_path: ./morphology_appropriate_representation_learning/preprocessed_data/other_polymorphic/manifest.csv
 model_checkpoints:
   [
-    "./morphology_appropriate_representation_learning/model_checkpoints/other_polymorphic/Rotation_invariant_pointcloud_SDF.ckpt",
-    "./morphology_appropriate_representation_learning/model_checkpoints/other_polymorphic/Rotation_invariant_image_SDF.ckpt",
+    "./morphology_appropriate_representation_learning/model_checkpoints/other_polymorphic/Classical_image_seg.ckpt",
     "./morphology_appropriate_representation_learning/model_checkpoints/other_polymorphic/Rotation_invariant_image_seg.ckpt",
     "./morphology_appropriate_representation_learning/model_checkpoints/other_polymorphic/Classical_image_SDF.ckpt",
-    "./morphology_appropriate_representation_learning/model_checkpoints/other_polymorphic/Classical_image_seg.ckpt",
+    "./morphology_appropriate_representation_learning/model_checkpoints/other_polymorphic/Rotation_invariant_image_SDF.ckpt",
+    "./morphology_appropriate_representation_learning/model_checkpoints/other_polymorphic/Rotation_invariant_pointcloud_SDF.ckpt",
   ]
 names:
   [
-    "Rotation_invariant_pointcloud_SDF",
-    "Rotation_invariant_image_SDF",
+    "Classical_image_seg",
     "Rotation_invariant_image_seg",
     "Classical_image_SDF",
-    "Classical_image_seg",
+    "Rotation_invariant_image_SDF",
+    "Rotation_invariant_pointcloud_SDF",
   ]
 data_paths:
   [
-    "/data/other_polymorphic/pc.yaml",
-    "/data/other_polymorphic/so3_image_sdf.yaml",
+    "/data/other_polymorphic/classical_image_seg.yaml",
     "/data/other_polymorphic/so3_image_seg.yaml",
     "/data/other_polymorphic/classical_image_sdf.yaml",
-    "/data/other_polymorphic/classical_image_seg.yaml",
+    "/data/other_polymorphic/so3_image_sdf.yaml",
+    "/data/other_polymorphic/pc.yaml",
   ]
 classification_label: ["structure_name"]
 regression_label:

docs/DEVELOPMENT.md

@@ -19,15 +19,16 @@ benchmarking_representations
 │   │   ├── cellpack
 │   │   ├── chandrasekaran_et_al
 │   │   ├── data
+│   │   │   ├── cellpack     <- Config files associated with cellpack simulations
 │   │   │   ├── get_datamodules.py     <- Get final list of datamodules per dataset
 │   │   │   └── preprocessing          <- Preprocessing scripts to generate point clouds and SDFs
-│   │   ├── features
 │   │   ├── features                   <- Metrics for benchmarking each model
 │   │   │   ├── archetype.py           <- Archetype analysis functions
 │   │   │   ├── classification.py      <- Test set classification accuracies using logistic regression classifiers
 │   │   │   ├── outlier_compactness.py <- Intrinsic dimensionality calculation and outlier classification
 │   │   │   ├── reconstruction.py      <- Functions for reconstruction viz across models
 │   │   │   ├── regression.py          <- Linear regression test set r^2
+│   │   │   ├── evolve.py          <- Evolution energy and interpolation distance metrics
 │   │   │   ├── rotation_invariance.py <- Sensitivity to four 90 degree rotations in embedding space
 │   │   │   └── plot.py                <- Polar plot viz across metrics
 │   │   ├── models                     <- Training and inference scripts
@@ -36,8 +37,9 @@ benchmarking_representations
 │   │   │   ├── save_embeddings.py     <- Save embeddings using inference functions
 │   │   │   ├── load_models.py         <- Load trained models based on checkpoint paths
 │   │   │   └── compute_features.py    <- Compute multi-metric features for each model based on saved embeddings
-│   │   ├── notebooks           <- Jupyter notebooks
 │   │   └── visualization
+│   │   │   ├── mitsuba_render_image.py               <- Mitsuba rendering for image segmentations
+│   │   │   ├── mitsuba_render_pc.py               <- Mitsuba rendering for pointclouds
 │   └── pointcloudutils
 │       ├── datamodules         <- Custom datamodules
 │       │   └── cellpack.py     <- CellPACK data specific datamodule

docs/PREPROCESSING.md

@@ -52,7 +52,8 @@ src
 
 # Polymorphic structures: Generate SDFs
 
-Use the segmentation data for polymorphic structures as input to the SDF generation step. 
+Use the segmentation data for polymorphic structures as input to the SDF generation step.
+
 ```
 src
 └── br