Minor touch-up to finetuning directory (#577)

Refactors the `finetuning` directory and moves old scripts under the `outdated` directory in the respective parent folders
computational-cell-analytics · May 2, 2024 · 79dff63 · 79dff63
1 parent 562bcb2
commit 79dff63
Show file tree

Hide file tree

Showing 44 changed files with 43 additions and 124 deletions.
diff --git a/finetuning/evaluation/mitonet_baseline.py b/finetuning/evaluation/mitonet_baseline.py
@@ -1,3 +1,9 @@
+# NOTE:
+# This script below the following:
+#  - stacks indivudual slice images into one volume for segmenting in empanada-napari
+#  - evaluates the segmentations with the mean segmentation accuracy metric
+
+
 import os
 from glob import glob
 

diff --git a/finetuning/generalists/README.md b/finetuning/generalists/README.md
@@ -0,0 +1,26 @@
+# Segment Anything Finetuning for Generalists Models
+
+Code for finetuning segment anything data on multiple microscopy datasets.
+
+## Finetuning Scripts
+
+- `training/`: The finetuning scripts for the two microscopy domains covered by `micro_sam`.
+    - Light Microscopy (LM)
+        - `light_microscopy/obtain_lm_datasets.py`: Scripts for getting the dataloader from multipe LM datasets.
+        - `light_microscopy/train_lm_generalist.py`: Finetuning on multiple LM datasets.
+    - Electron Microscopy (EM)
+        - `mito_nuc/obtain_mito_nuc_em_datasets.py`: Scripts for getting the dataloader from multiple EM datasets.
+        - `mito_nuc/train_mito_nuc_em_generalist.py`: Finetuning on multiple EM datasets for segmenting mitochondria and nuclei.
+
+
+## Experimental Scripts
+These scripts are a work-in-progress and often under active development.
+
+- `training/histopathology`: Finetuning Segment Anything on Histopathology datasets.
+- `training/electron_microscopy/boundaries`: Finetuning Segment Anything on Electron Microscopy datasets for segmenting boundary-based dense structures.
+
+> For details on how to run the scripts from above: `python <DATASET>_finetuning.py -h`
+
+
+## Outdated Scripts
+The scripts located at `outdated/` are not in working purpose with the latest version of `micro-sam`.
diff --git a/finetuning/generalists/export_generalist_model.py b/finetuning/generalists/export_generalist_model.py
diff --git a/finetuning/generalists/generate_model_comparison.py b/finetuning/generalists/generate_model_comparison.py
diff --git a/finetuning/generalists/compile_results.py → ...g/generalists/outdated/compile_results.py b/finetuning/generalists/compile_results.py → ...g/generalists/outdated/compile_results.py
diff --git a/...ning/generalists/create_tissuenet_data.py → ...ralists/outdated/create_tissuenet_data.py b/...ning/generalists/create_tissuenet_data.py → ...ralists/outdated/create_tissuenet_data.py
diff --git a/...tuning/generalists/evaluate_generalist.py → ...neralists/outdated/evaluate_generalist.py b/...tuning/generalists/evaluate_generalist.py → ...neralists/outdated/evaluate_generalist.py
diff --git a/...ng/generalists/evaluate_generalist.sbatch → ...lists/outdated/evaluate_generalist.sbatch b/...ng/generalists/evaluate_generalist.sbatch → ...lists/outdated/evaluate_generalist.sbatch
diff --git a/...eneralists/evaluate_training_evolution.py → ...s/outdated/evaluate_training_evolution.py b/...eneralists/evaluate_training_evolution.py → ...s/outdated/evaluate_training_evolution.py
diff --git a/...alists/evaluate_training_evolution.sbatch → ...tdated/evaluate_training_evolution.sbatch b/...alists/evaluate_training_evolution.sbatch → ...tdated/evaluate_training_evolution.sbatch
diff --git a/finetuning/generalists/precompute_prompts.py → ...eneralists/outdated/precompute_prompts.py b/finetuning/generalists/precompute_prompts.py → ...eneralists/outdated/precompute_prompts.py
diff --git a/finetuning/generalists/util.py → finetuning/generalists/outdated/util.py b/finetuning/generalists/util.py → finetuning/generalists/outdated/util.py
diff --git a/finetuning/generalists/precompute_prompts.sbatch b/finetuning/generalists/precompute_prompts.sbatch
diff --git a/finetuning/specialists/README.md b/finetuning/specialists/README.md
@@ -1,4 +1,4 @@
-# Segment Anything Finetuning for Specific Datasets
+# Segment Anything Finetuning for Specialist Models
 
 Code for finetuning segment anything data on specific microscopy datasets.
 
@@ -17,3 +17,8 @@ Code for finetuning segment anything data on specific microscopy datasets.
 > For details on how to run the scripts from above: `python <DATASET>_finetuning.py -h`
 
 - `resource_efficient_finetuning`: The experiments for finetuning a custom dataset on limited resources.
+
+
+## Outdated Scripts
+The scripts located at `outdated/` are not in working purpose with the latest version of `micro-sam`.
+- It comprises of extensive experiments on "LIVECell" specialist, located at `outdated/livecell/`.
diff --git a/finetuning/livecell/README.md → ...g/specialists/outdated/livecell/README.md b/finetuning/livecell/README.md → ...g/specialists/outdated/livecell/README.md
@@ -1,16 +1,14 @@
-# Finetuning Segment Anything for LiveCELL
+# Finetuning Segment Anything for LIVECell
 
-TODO: explain the set-up
-
-These experiments are implemented for a slurm cluster with access to GPUs (and you ideally need access to A100s or H100s with 80GB of memory, if you only use ViT-b then using a GPU with 32 GB or 40 GB should suffice.)
+These experiments are implemented for a slurm cluster with access to GPUs (and you ideally need access to A100s or H100s with 80GB of memory, if you only use ViT Base then using a GPU with 32 GB or 40 GB should suffice.)
 
 ## Training
 
-TODO: add training code and explain how to run it
+The relevant scripts are located in the top-level `finetuning` directory at: `finetuning/livecell_finetuning.py`.
 
 ## Evaluation
 
-To run the evaluation experiments for the Segment Anything Models on LiveCELL follow these steps:
+To run the evaluation experiments for the Segment Anything Models on LIVECell follow these steps:
 
 - Preparation:
     - Go to the `evaluation` directory.

diff --git a/...uning/livecell/evaluation/evaluate_amg.py → ...dated/livecell/evaluation/evaluate_amg.py b/...uning/livecell/evaluation/evaluate_amg.py → ...dated/livecell/evaluation/evaluate_amg.py
diff --git a/...g/livecell/evaluation/evaluate_amg.sbatch → ...d/livecell/evaluation/evaluate_amg.sbatch b/...g/livecell/evaluation/evaluate_amg.sbatch → ...d/livecell/evaluation/evaluate_amg.sbatch
diff --git a/...luation/evaluate_instance_segmentation.py → ...luation/evaluate_instance_segmentation.py b/...luation/evaluate_instance_segmentation.py → ...luation/evaluate_instance_segmentation.py
diff --git a/...ion/evaluate_instance_segmentation.sbatch → ...ion/evaluate_instance_segmentation.sbatch b/...ion/evaluate_instance_segmentation.sbatch → ...ion/evaluate_instance_segmentation.sbatch
diff --git a/...ivecell/evaluation/iterative_prompting.py → ...ivecell/evaluation/iterative_prompting.py b/...ivecell/evaluation/iterative_prompting.py → ...ivecell/evaluation/iterative_prompting.py
diff --git a/...ell/evaluation/iterative_prompting.sbatch → ...ell/evaluation/iterative_prompting.sbatch b/...ell/evaluation/iterative_prompting.sbatch → ...ell/evaluation/iterative_prompting.sbatch
diff --git a/...ing/livecell/evaluation/old/evaluation.py → ...ted/livecell/evaluation/old/evaluation.py b/...ing/livecell/evaluation/old/evaluation.py → ...ted/livecell/evaluation/old/evaluation.py
diff --git a/...livecell/evaluation/old/evaluation.sbatch → ...livecell/evaluation/old/evaluation.sbatch b/...livecell/evaluation/old/evaluation.sbatch → ...livecell/evaluation/old/evaluation.sbatch
diff --git a/...ning/livecell/evaluation/old/inference.py → ...ated/livecell/evaluation/old/inference.py b/...ning/livecell/evaluation/old/inference.py → ...ated/livecell/evaluation/old/inference.py
diff --git a/.../livecell/evaluation/old/inference.sbatch → .../livecell/evaluation/old/inference.sbatch b/.../livecell/evaluation/old/inference.sbatch → .../livecell/evaluation/old/inference.sbatch
diff --git a/...cell/evaluation/old/precompute_prompts.py → ...cell/evaluation/old/precompute_prompts.py b/...cell/evaluation/old/precompute_prompts.py → ...cell/evaluation/old/precompute_prompts.py
diff --git a/.../evaluation/old/precompute_prompts.sbatch → .../evaluation/old/precompute_prompts.sbatch b/.../evaluation/old/precompute_prompts.sbatch → .../evaluation/old/precompute_prompts.sbatch
diff --git a/...ecell/evaluation/precompute_embeddings.py → ...ecell/evaluation/precompute_embeddings.py b/...ecell/evaluation/precompute_embeddings.py → ...ecell/evaluation/precompute_embeddings.py
diff --git a/...l/evaluation/precompute_embeddings.sbatch → ...l/evaluation/precompute_embeddings.sbatch b/...l/evaluation/precompute_embeddings.sbatch → ...l/evaluation/precompute_embeddings.sbatch
diff --git a/.../livecell/evaluation/submit_evaluation.py → .../livecell/evaluation/submit_evaluation.py b/.../livecell/evaluation/submit_evaluation.py → .../livecell/evaluation/submit_evaluation.py
diff --git a/...cell/evaluation/train_grid_search_eval.py → ...cell/evaluation/train_grid_search_eval.py b/...cell/evaluation/train_grid_search_eval.py → ...cell/evaluation/train_grid_search_eval.py
diff --git a/finetuning/livecell/evaluation/util.py → ...ists/outdated/livecell/evaluation/util.py b/finetuning/livecell/evaluation/util.py → ...ists/outdated/livecell/evaluation/util.py
diff --git a/finetuning/livecell/experiments/README.md → ...s/outdated/livecell/experiments/README.md b/finetuning/livecell/experiments/README.md → ...s/outdated/livecell/experiments/README.md
diff --git a/.../experiments/run_experiment_evaluation.py → .../experiments/run_experiment_evaluation.py b/.../experiments/run_experiment_evaluation.py → .../experiments/run_experiment_evaluation.py
diff --git a/...periments/submit_experiment_finetuning.py → ...periments/submit_experiment_finetuning.py b/...periments/submit_experiment_finetuning.py → ...periments/submit_experiment_finetuning.py
diff --git a/...ecell/joint_training/grid_search_train.py → ...ecell/joint_training/grid_search_train.py b/...ecell/joint_training/grid_search_train.py → ...ecell/joint_training/grid_search_train.py
diff --git a/...vecell/joint_training/joint_finetuning.py → ...vecell/joint_training/joint_finetuning.py b/...vecell/joint_training/joint_finetuning.py → ...vecell/joint_training/joint_finetuning.py
diff --git a/...tuning/livecell/joint_training/results.md → ...tdated/livecell/joint_training/results.md b/...tuning/livecell/joint_training/results.md → ...tdated/livecell/joint_training/results.md
diff --git a/...oint_training/run_parameter_evaluation.py → ...oint_training/run_parameter_evaluation.py b/...oint_training/run_parameter_evaluation.py → ...oint_training/run_parameter_evaluation.py
diff --git a/...ivecell/joint_training/unetr_inference.py → ...ivecell/joint_training/unetr_inference.py b/...ivecell/joint_training/unetr_inference.py → ...ivecell/joint_training/unetr_inference.py
diff --git a/...inetuning/evaluate_partially_finetuned.py → ...inetuning/evaluate_partially_finetuned.py b/...inetuning/evaluate_partially_finetuned.py → ...inetuning/evaluate_partially_finetuned.py
diff --git a/...uning/evaluate_partially_finetuned.sbatch → ...uning/evaluate_partially_finetuned.sbatch b/...uning/evaluate_partially_finetuned.sbatch → ...uning/evaluate_partially_finetuned.sbatch
diff --git a/...l_finetuning/submit_partial_finetuning.py → ...l_finetuning/submit_partial_finetuning.py b/...l_finetuning/submit_partial_finetuning.py → ...l_finetuning/submit_partial_finetuning.py
diff --git a/finetuning/specialists/resource-efficient/README.md b/finetuning/specialists/resource-efficient/README.md
@@ -40,7 +40,7 @@ Fixed parameters:
 
 ### GPU Resources
 
-(32G CPU memory, 8 CPU cores)
+(32GB CPU memory, 8 CPU cores)
 
 1. `gtx1080`:
     - `vit_t`: finetune all layers