Moseq Pipeline (LorenFrankLab#1056)

* Add organization tables for keypoint pose data * Add model training tables * apply fit moseq model to data * add ability to initialize model training from pre-existing model * make model names unique * add initial tutorial * add moseq pipeline diagram * get rid of PoseOutput and fetch pose data from PositionOutput * remove PoseOutput and corresponding references * allow passing of null entries from DLCPoseEstimation to PositionOutput for Pose-dependent analysis * Apply suggestions from code review Co-authored-by: Chris Broz <[email protected]> * lint * implement changes from review * rename fetch_video_name to fetch_video_path * cleanup outdated usages of PoseOutput * Cleanup errors from removal of PoseOutput * add get_position_interval_epoch * Add method to get training results pdf * Add description and example of hyperparameter sweep * move moseq dir definition to spyglass config * move moseq config function to method * Update changelog * update pipeline diagram * Apply suggestions from code review Co-authored-by: Chris Broz <[email protected]> * move moseq into v1 folder * add docstrings * make video symlink more robust * cleanup setup_project call * cleanup config method * Implement suggestions from code review * cleanup readability of DLCPosV1 make conditions * allow key argument when fetching video path * add moseq dependencies * Add moseq install instructions to tutorials * accept key in all get video path functions * fix spelling * fix spelling * make model project dir unique * add default empty restriction * add moseq to omitted tests * tutorial error cleanup * fix for codespell update * typo fix * remove jax version pin * add docstrings * make restriction in fetch_pose_dataframe compatible with export * add moseq dirs to config example * Update src/spyglass/position/v1/position_dlc_pose_estimation.py Co-authored-by: Chris Broz <[email protected]> * cleanup long settings if statements * change to and register the behavior schema prefix * Update src/spyglass/position/position_merge.py Co-authored-by: Chris Broz <[email protected]> * lint --------- Co-authored-by: Samuel Bray <[email protected]> Co-authored-by: Samuel Bray <[email protected]> Co-authored-by: Chris Broz <[email protected]>
CBroz1 · Feb 27, 2025 · 2b71fb3 · 2b71fb3
1 parent 4f39dfa
commit 2b71fb3
Show file tree

Hide file tree

Showing 26 changed files with 4,245 additions and 57 deletions.
diff --git a/CHANGELOG.md b/CHANGELOG.md
@@ -6,6 +6,7 @@
 
 - Ensure merge tables are declared during file insertion #1205
 - Update URL for DANDI Docs #1210
+- Add common method `get_position_interval_epoch` #1056
 - Improve cron job documentation and script #1226, #1241
 - Update export process to include `~external` tables #1239
 - Only add merge parts to `source_class_dict` if present in codebase #1237
@@ -22,6 +23,10 @@
     - Fix compatibility bug between v1 pipeline and `SortedSpikesGroup` unit
         filtering #1238
 
+- Behavior
+    - Implement pipeline for keypoint-moseq extraction of behavior syllables #1056
+
+
 ## [0.5.4] (December 20, 2024)
 
 ### Infrastructure

diff --git a/dj_local_conf_example.json b/dj_local_conf_example.json
@@ -51,6 +51,11 @@
       "video": "/your/base/path/deeplabcut/video",
       "output": "/your/base/path/deeplabcut/output"
     },
+    "moseq_dirs": {
+      "base": "/your/base/path/moseq",
+      "project": "/your/base/path/moseq/projects",
+      "video": "/your/base/path/moseq/video",
+    },
     "kachery_zone": "franklab.default"
   }
 }
diff --git a/environment_moseq_cpu.yml b/environment_moseq_cpu.yml
@@ -0,0 +1,42 @@
+# 1. Install a conda distribution.
+#    https://mamba.readthedocs.io/en/latest/installation/mamba-installation.html
+# 2. Run: `mamba env create -f environment.yml`
+# 3. Activate: `conda activate spyglass`
+#
+# (lines intentionally left blank)
+#
+#
+name: spyglass-moseq-cpu
+channels:
+  - conda-forge
+  # - defaults # deprecated
+  - franklab
+  - edeno
+  # - pytorch # dlc-only
+  # - anaconda  # dlc-only, for cudatoolkit
+dependencies:
+  - bottleneck
+  # - cudatoolkit=11.3 # dlc-only
+  # - ffmpeg # dlc-only
+  - ipympl
+  - jupyterlab>=3.*
+  # - libgcc # dlc-only
+  - matplotlib
+  - non_local_detector
+  - numpy
+  - pip
+  - position_tools
+  - pybind11  # req by mountainsort4 -> isosplit5
+  - pydotplus
+  - pyfftw<=0.12.0  # ghostipy req. install from conda-forge for Mac ARM
+  - python>=3.9,<3.13
+  - pytorch<1.12.0
+  - ripple_detection
+  - seaborn
+  # - torchaudio # dlc-only
+  # - torchvision # dlc-only
+  - track_linearization>=2.3
+  - pip:
+      - ghostipy # for common_filter
+      - mountainsort4
+      - .[moseq-cpu]
diff --git a/environment_moseq_gpu.yml b/environment_moseq_gpu.yml
@@ -0,0 +1,42 @@
+# 1. Install a conda distribution.
+#    https://mamba.readthedocs.io/en/latest/installation/mamba-installation.html
+# 2. Run: `mamba env create -f environment.yml`
+# 3. Activate: `conda activate spyglass`
+#
+# (lines intentionally left blank)
+#
+#
+name: spyglass-moseq-gpu
+channels:
+  - conda-forge
+  # - defaults # deprecated
+  - franklab
+  - edeno
+  # - pytorch # dlc-only
+  # - anaconda  # dlc-only, for cudatoolkit
+dependencies:
+  - bottleneck
+  # - cudatoolkit=11.3 # dlc-only
+  # - ffmpeg # dlc-only
+  - ipympl
+  - jupyterlab>=3.*
+  # - libgcc # dlc-only
+  - matplotlib
+  - non_local_detector
+  - numpy
+  - pip
+  - position_tools
+  - pybind11  # req by mountainsort4 -> isosplit5
+  - pydotplus
+  - pyfftw<=0.12.0  # ghostipy req. install from conda-forge for Mac ARM
+  - python>=3.9,<3.13
+  - pytorch<1.12.0
+  - ripple_detection
+  - seaborn
+  # - torchaudio # dlc-only
+  # - torchvision # dlc-only
+  - track_linearization>=2.3
+  - pip:
+      - ghostipy # for common_filter
+      - mountainsort4
+      - .[moseq-gpu]
diff --git a/notebook-images/moseq_outline.png b/notebook-images/moseq_outline.png
diff --git a/notebooks/00_Setup.ipynb b/notebooks/00_Setup.ipynb
@@ -285,7 +285,47 @@
     "The Decoding pipeline relies on `jax` to process data with GPUs. Please see\n",
     "their conda installation steps\n",
     "[here](https://jax.readthedocs.io/en/latest/installation.html#conda-installation).\n",
-    "\n"
+    "\n",
+    "#### Deep Lab Cut (DLC)\n",
+    "\n",
+    "Spyglass provides an environment build for using the DLC pipeline. To create an \n",
+    "environment with these features, please:\n",
+    "1. navigate to your cloned spyglass repo.\n",
+    "2. build the environment from the dlc version\n",
+    "3. activate the environment to use\n",
+    "\n",
+    "```bash\n",
+    "cd /path/to/spyglass # 1\n",
+    "mamba env create -f environment_dlc.yml # 2\n",
+    "mamba activate spyglass-dlc # 3\n",
+    "```\n",
+    "\n",
+    "Alternatively, you can pip install using\n",
+    "```bash\n",
+    "pip install spyglass[dlc]\n",
+    "```\n",
+    "\n",
+    "#### Keypoint-Moseq\n",
+    "\n",
+    "Spyglass provides an environment build for using the Moseq pipeline. To create an \n",
+    "environment with these features, please:\n",
+    "1. navigate to your cloned spyglass repo.\n",
+    "2. build the environment from one of the moseq versions\n",
+    "3. activate the environment to use\n",
+    "\n",
+    "```bash\n",
+    "cd /path/to/spyglass # 1\n",
+    "mamba env create -f environment_moseq_cpu.yml # 2\n",
+    "mamba activate spyglass-moseq-cpu # 3\n",
+    "```\n",
+    "\n",
+    "Alternatively, you can pip install using\n",
+    "```bash\n",
+    "pip install spyglass[moseq-cpu]\n",
+    "```\n",
+    "\n",
+    "To use a GPU enabled version of the package, replace `cpu` with `gpu` in the above \n",
+    "commands\n"
    ]
   },
   {
@@ -593,7 +633,7 @@
  ],
  "metadata": {
   "kernelspec": {
-   "display_name": "Python 3 (ipykernel)",
+   "display_name": "spyglass",
    "language": "python",
    "name": "python3"
   },
@@ -607,7 +647,7 @@
    "name": "python",
    "nbconvert_exporter": "python",
    "pygments_lexer": "ipython3",
-   "version": "3.9.19"
+   "version": "3.9.18"
   }
  },
  "nbformat": 4,