Add codespell support (config, workflow to detect/not fix) and make it fix few typos

.codespellrc

@@ -0,0 +1,6 @@
+[codespell]
+# Ref: https://github.com/codespell-project/codespell#using-a-config-file
+skip = .git*,.codespellrc,FSharp.Core.xml
+check-hidden = true
+# ignore-regex = 
+ignore-words-list = nd,ans

.github/workflows/codespell.yml

@@ -0,0 +1,25 @@
+# Codespell configuration is within .codespellrc
+---
+name: Codespell
+
+on:
+  push:
+    branches: [master]
+  pull_request:
+    branches: [master]
+
+permissions:
+  contents: read
+
+jobs:
+  codespell:
+    name: Check for spelling errors
+    runs-on: ubuntu-latest
+
+    steps:
+      - name: Checkout
+        uses: actions/checkout@v4
+      - name: Annotate locations with typos
+        uses: codespell-project/codespell-problem-matcher@v1
+      - name: Codespell
+        uses: codespell-project/actions-codespell@v2

README.md

@@ -52,7 +52,7 @@ cd dannce
 Then you should be ready to try the quickstart demo! \
 These installation steps were tested with Anaconda releases 4.7.12 and 2020.02, although we expect it to work for most conda installations. After installing Anaconda, and presuming there are no issues with GPU drivers, the installation should take less than 5 minutes.
 
-A note on the PyTorch requirment.
+A note on the PyTorch requirement.
 PyTorch is not required, but 3D volume generation is significantly faster when using PyTorch than with TensorFlow or NumPy. To use TensorFlow only, without having to install the PyTorch package, simply toggle the `predict_mode` field in the DANNCE configuration files to `tf`. To use NumPy volume generation (slowest), change `predict_mode` to `None`.
 
 
@@ -149,7 +149,7 @@ DANNCE requires a parent video directory with *n* sub-directories, one for each
 `DANNCE` uses two configuration files and one data file. 
 
 - *main config*, e.g. `configs/*.yaml`. This file defines data and model hyperparameters. It can be reused across experiments.  
-- *io config*, e.g. `demo/markerless_mouse_1/io.yaml`. This file defines input data and ouput directories. It is used for a single experiment. 
+- *io config*, e.g. `demo/markerless_mouse_1/io.yaml`. This file defines input data and output directories. It is used for a single experiment. 
 - *dannce.mat*, e.g. `demo/markerless_mouse_1/label3d_dannce.mat`. This file contains three cell arrays of matlab structures. `params` stores the camera parameters for each camera. `sync` stores a vector that synchronizes all cameras. `labelData` stores the frame identities and 3d labels for hand-labeled frames. This file can be produced automatically with `Label3D.exportDannce()`.
 
 **camera calibration parameters**.
@@ -160,7 +160,7 @@ A properly formatted calibration struct has the following fields, `['R','t','K',
 **synchronization files**.
 DANNCE requires a set of sync structs, one for each camera, which define frame synchrony across the different cameras over time. If you know your cameras are reliably synchronized at all times (e.g. via hardware triggering), these files can be generated with the aid of `dannce/utils/makeSyncFiles.py`. Once your video directories are set up correctly, sync files can get generated by running `python dannce/utils/makeSyncFiles.py {path_to_videos} {acquisition_frame_rate} {number_tracked_landmarks}`, where {.} denotes variables you must replace with relevant values. See the `makeSyncFiles.py` docstring for more information.
 
-If your cameras are not natively synchronized, but you can collect timestaps for each frame, sync files should be generated by `dannce/utils/preprocess_data.m`, which will generate sync files from a properly formatted `.mat` file listing the frameID for each camera at each timepoint. See `/dannce/utils/example_matchedframs.mat` file for how these timestamp data should be formatted before running `preprocess_data.m`.
+If your cameras are not natively synchronized, but you can collect timestamps for each frame, sync files should be generated by `dannce/utils/preprocess_data.m`, which will generate sync files from a properly formatted `.mat` file listing the frameID for each camera at each timepoint. See `/dannce/utils/example_matchedframs.mat` file for how these timestamp data should be formatted before running `preprocess_data.m`.
 
 ## Hand-Labeling
 For fine-tuning DANNCE to work with your animal and system, we developed a labeling GUI, which can be found in a separate repo: https://github.com/diegoaldarondo/Label3D. The `Label3D` repository should be cloned with DANNCE automatically as a submodule when using `git clone --recursive https://github.com/spoonsso/dannce` When labeling is completed, the labels can be used to train DANNCE and the COMfinder network (see below) after converting the Label3D files to DANNCE format using `Label3D.exportDannce()`.

calibration/acquire_calibration_3cam_mouse_clean.m

@@ -56,7 +56,7 @@
 %preview(vid{1});
 
 
-%% get matched frames from the cameras wth
+%% get matched frames from the cameras with
 %% STEP 1- CHECKERBOARD
 fprintf('starting in 10 s')
 pause(10)
@@ -73,7 +73,7 @@
     pause(0.5)
 end
 
-%% acquire and save postion of markers
+%% acquire and save position of markers
 %% STEP 2 (Put L-frame in arena first)
 
 input('Hit enter when the grid is in the arena \n')
@@ -207,7 +207,7 @@
     worldLocation = cell(1,numcams);
     rotationMatrix = cell(1,numcams);
     translationVector = cell(1,numcams);
-    %% get the orienation and location of cameras
+    %% get the orientation and location of cameras
     for kk = [1:numcams]%:numcams
 
      [worldOrientation{kk},worldLocation{kk}] = estimateWorldCameraPose(double(point_coordinates{kk}),...

calibration/multi-camera calibration/calibration_extrinsic_Lframe.m

@@ -65,7 +65,7 @@
 %% Use selected points to calculate camera extrinsics
 for kk = 1:numel(lframe)
     %kk
-    % Do grid search over parms
+    % Do grid search over params
     curr_err = 1e10;
     c_save = 0;
     mr_save = 0;

configs/dannce_rig_dannce_config.yaml

@@ -86,10 +86,10 @@ depth: False
 
 immode: 'vid'
 
-# DANNCE training option. Whetehr to turn on rotation augmentation during training
+# DANNCE training option. Whether to turn on rotation augmentation during training
 rotate: True
 
-# If true, intializes an "AVG" version of the network (i.e. final spatial expected value output layer). If false, "MAX" version
+# If true, initializes an "AVG" version of the network (i.e. final spatial expected value output layer). If false, "MAX" version
 expval: True
 
 

dannce/cli.py

@@ -261,7 +261,7 @@ def add_shared_train_args(
         "--data-split-seed",
         dest="data_split_seed",
         type=int,
-        help="Integer seed for the random numebr generator controlling train/test data splits",
+        help="Integer seed for the random number generator controlling train/test data splits",
     )
     parser.add_argument(
         "--valid-exp",

dannce/engine/generator.py

@@ -1066,7 +1066,7 @@ def pj_grid_mirror(self, X_grid, camname, ID, experimentID, thisim):
 
     def pj_grid_post(self, X_grid, camname, ID, experimentID,
                      com, com_precrop, thisim):
-        # separate the porjection and sampling into its own function so that
+        # separate the projection and sampling into its own function so that
         # when mirror == True, this can be called directly
         if self.crop_im:
             if self.torch.all(self.torch.isnan(com)):

dannce/engine/generator_aux.py

@@ -142,7 +142,7 @@ def __data_generation(self, list_IDs_temp):
             )
 
         # We'll need to transpose this later such that channels are last,
-        # but initializaing the array this ways gives us
+        # but initializing the array this ways gives us
         # more flexibility in terms of user-defined array sizes\
         if self.labelmode == "prob":
             y = np.empty(
@@ -244,7 +244,7 @@ def __data_generation(self, list_IDs_temp):
             y = np.transpose(y, [0, 2, 3, 1])
 
             if self.mirror:
-                # separate the batches from the cameras, and use the cameras as the numebr of channels 
+                # separate the batches from the cameras, and use the cameras as the number of channels 
                 # to make a single-shot multi-target prediction from a single image
                 y = np.reshape(y, (self.batch_size, len(self.camnames[0]), y.shape[1], y.shape[2]))
                 y = np.transpose(y, [0, 2, 3, 1])
@@ -393,7 +393,7 @@ def __data_generation(self, list_IDs_temp):
         )
 
         # We'll need to transpose this later such that channels are last,
-        # but initializaing the array this ways gives us
+        # but initializing the array this ways gives us
         # more flexibility in terms of user-defined array sizes\
         if self.labelmode == "prob":
             y = np.empty(

dannce/engine/inference.py

@@ -373,10 +373,10 @@ def triangulate_single_instance(
     """Triangulate for a single instance.
 
     Args:
-        n_cams (int): Numver of cameras
+        n_cams (int): Number of cameras
         sample_id (Text): Sample identifier.
         params (Dict): Parameters dictionary.
-        camera_mats (Dict): Camera matrices dictioanry.
+        camera_mats (Dict): Camera matrices dictionary.
         save_data (Dict): Saved data dictionary.
 
     No Longer Returned:
@@ -409,10 +409,10 @@ def triangulate_multi_instance_multi_channel(
     """Triangulate for multi-instance multi-channel.
 
     Args:
-        n_cams (int): Numver of cameras
+        n_cams (int): Number of cameras
         sample_id (Text): Sample identifier.
         params (Dict): Parameters dictionary.
-        camera_mats (Dict): Camera matrices dictioanry.
+        camera_mats (Dict): Camera matrices dictionary.
         save_data (Dict): Saved data dictionary.
 
     No Longer Returned:
@@ -467,10 +467,10 @@ def triangulate_multi_instance_single_channel(
     """Triangulate for multi-instance single-channel.
 
     Args:
-        n_cams (int): Numver of cameras
+        n_cams (int): Number of cameras
         sample_id (Text): Sample identifier.
         params (Dict): Parameters dictionary.
-        camera_mats (Dict): Camera matrices dictioanry.
+        camera_mats (Dict): Camera matrices dictionary.
         cameras (Dict): Camera dictionary.
         save_data (Dict): Saved data dictionary.
 

dannce/engine/losses.py

@@ -19,7 +19,7 @@ def mask_nan(y_true, y_pred):
 
 
 def mask_nan_keep_loss(y_true, y_pred):
-    """Mask out nan values in the calulation of MSE."""
+    """Mask out nan values in the calculation of MSE."""
     y_pred, y_true, num_notnan = mask_nan(y_true, y_pred)
     loss = K.sum((K.flatten(y_pred) - K.flatten(y_true)) ** 2) / num_notnan
     return tf.where(~tf.math.is_nan(loss), loss, 0)
@@ -53,7 +53,7 @@ def metric_dist_max(y_true, y_pred):
     """Get distance between the (row, col) indices of each maximum.
 
     y_true and y_pred are image-sized confidence maps.
-    Let's get the (row, col) indicies of each maximum and calculate the
+    Let's get the (row, col) indices of each maximum and calculate the
     distance between the two
     """
     x = K.reshape(y_true, [K.int_shape(y_true)[0], -1])

dannce/engine/ops.py

@@ -18,7 +18,7 @@ def camera_matrix(K: np.ndarray, R: np.ndarray, t: np.ndarray) -> np.ndarray:
     """Derive the camera matrix.
 
     Derive the camera matrix from the camera intrinsic matrix (K),
-    and the extrinsic rotation matric (R), and extrinsic
+    and the extrinsic rotation matrix (R), and extrinsic
     translation vector (t).
 
     Note that this uses the matlab convention, such that
@@ -33,7 +33,7 @@ def project_to2d(
     """Project 3d points to 2d.
 
     Projects a set of 3-D points, pts, into 2-D using the camera intrinsic
-    matrix (K), and the extrinsic rotation matric (R), and extrinsic
+    matrix (K), and the extrinsic rotation matrix (R), and extrinsic
     translation vector (t). Note that this uses the matlab
     convention, such that
     M = [R;t] * K, and pts2d = pts3d * M
@@ -50,7 +50,7 @@ def project_to2d_torch(pts, M: np.ndarray, device: Text) -> torch.Tensor:
     """Project 3d points to 2d.
 
     Projects a set of 3-D points, pts, into 2-D using the camera intrinsic
-    matrix (K), and the extrinsic rotation matric (R), and extrinsic
+    matrix (K), and the extrinsic rotation matrix (R), and extrinsic
     translation vector (t). Note that this uses the matlab
     convention, such that
     M = [R;t] * K, and pts2d = pts3d * M
@@ -71,7 +71,7 @@ def project_to2d_tf(projPts, M):
     """Project 3d points to 2d.
 
     Projects a set of 3-D points, pts, into 2-D using the camera intrinsic
-    matrix (K), and the extrinsic rotation matric (R), and extrinsic
+    matrix (K), and the extrinsic rotation matrix (R), and extrinsic
     translation vector (t). Note that this uses the matlab
     convention, such that
     M = [R;t] * K, and pts2d = pts3d * M
@@ -561,7 +561,7 @@ def unDistortPoints(
 def triangulate(pts1, pts2, cam1, cam2):
     """Return triangulated 3- coordinates.
 
-    Following Matlab convetion, given lists of matching points, and their
+    Following Matlab convention, given lists of matching points, and their
     respective camera matrices, returns the triangulated 3- coordinates.
     pts1 and pts2 must be Mx2, where M is the number of points with
     (x,y) positions. M 3-D points will be returned after triangulation
@@ -599,7 +599,7 @@ def triangulate(pts1, pts2, cam1, cam2):
 def triangulate_multi_instance(pts, cams):
     """Return triangulated 3- coordinates.
 
-    Following Matlab convetion, given lists of matching points, and their
+    Following Matlab convention, given lists of matching points, and their
     respective camera matrices, returns the triangulated 3- coordinates.
     pts1 and pts2 must be Mx2, where M is the number of points with
     (x,y) positions. M 3-D points will be returned after triangulation
@@ -754,7 +754,7 @@ def proj_slice(
     # x,y,z position. In my case, there should only be as many grids as there
     # are samples in the mini-batch,
     # but for some reason this code allows multiple 3D grids per sample.
-    # the order in rows (for the last 3 cols) should be rougly like this:
+    # the order in rows (for the last 3 cols) should be roughly like this:
     # [batch1_grid1_allcam1samples_locs, batch1_grid1_allcam2sample_locs,
     # batch1_grid1_allcam3sample_locs, batch1_grid2_allcam1samples_locs, ...]
     g_val = nearest3(sample_grid, sample_idx, clip=True)

dannce/engine/processing.py

@@ -559,7 +559,7 @@ def prepare_save_metadata(params):
         the 'experiment' field
     """
 
-    # Need to convert None to string but still want to conserve the metadat structure
+    # Need to convert None to string but still want to conserve the metadata structure
     # format, so we don't want to convert the whole dict to a string
     meta = params.copy()
 
@@ -1215,7 +1215,7 @@ def savedata_tomat(
 
     if addCOM is not None:
         # We use the passed comdict to add back in the com, this is useful
-        # if one wnats to bootstrap on these values for COMnet or otherwise
+        # if one wants to bootstrap on these values for COMnet or otherwise
         for i in range(len(sID)):
             pred_out_world[i] = pred_out_world[i] + addCOM[int(sID)][:, np.newaxis]
 

dannce/engine/serve_data_DANNCE.py

@@ -354,7 +354,7 @@ def prepare_COM(
                 elif method == "median":
                     com3d = np.nanmedian(com3d, axis=1)
                 else:
-                    raise Exception("Uknown 3D COM method")
+                    raise Exception("Unknown 3D COM method")
 
             com3d_dict[key] = com3d
         else:
@@ -397,7 +397,7 @@ def remove_samples(s, d3d, mode="clean", auxmode=None):
     mode == 'SpineM' means only remove data where SpineM is missing
     mode == 'liberal' means include any data that isn't *all* nan
     aucmode == 'JDM52d2' removes a really bad marker period -- samples 20k to 32k
-    I need to cull the samples array (as this is used to index eveyrthing else),
+    I need to cull the samples array (as this is used to index everything else),
     but also the
     data_3d_ array that is used to for finding clusters
     """

dannce/interface.py

@@ -590,7 +590,7 @@ def com_train(params: Dict):
     )
 
     def write_debug(trainData=True):
-        """Factoring re-used debug output code.
+        """Factoring reused debug output code.
 
         Writes training or validation images to an output directory, together
         with the ground truth COM labels and predicted COM labels, respectively.
@@ -687,7 +687,7 @@ def dannce_train(params: Dict):
     params["net"] = getattr(nets, params["net"])
 
     # Default to 6 views but a smaller number of views can be specified in the
-    # DANNCE config. If the legnth of the camera files list is smaller than
+    # DANNCE config. If the length of the camera files list is smaller than
     # n_views, relevant lists will be duplicated in order to match n_views, if
     # possible.
     n_views = int(params["n_views"])
@@ -1276,7 +1276,7 @@ def dannce_predict(params: Dict):
     """Predict with dannce network
 
     Args:
-        params (Dict): Paremeters dictionary.
+        params (Dict): Parameters dictionary.
     """
     # Depth disabled until next release.
     params["depth"] = False
@@ -1291,7 +1291,7 @@ def dannce_predict(params: Dict):
     netname = params["net"]
     params["net"] = getattr(nets, params["net"])
     # Default to 6 views but a smaller number of views can be specified in the DANNCE config.
-    # If the legnth of the camera files list is smaller than n_views, relevant lists will be
+    # If the length of the camera files list is smaller than n_views, relevant lists will be
     # duplicated in order to match n_views, if possible.
     n_views = int(params["n_views"])
 

dannce/utils/makeStructuredData.py

@@ -169,7 +169,7 @@
                 # faster intersections are tricky given that matched_frames are repeated multiple times
                 # another shortcut we are taking here is that we take a single mocap value rather than an average
                 # over all samples for a given frame
-                # Another problem is that the mframes might jump from oen file to the next, leaving a big gap
+                # Another problem is that the mframes might jump from one file to the next, leaving a big gap
                 # relative to predictions
                 dfsi = np.logical_and(dframe >= np.min(mframes), dframe <= np.max(mframes))
                 dframe_sub = dframe[dfsi]

dannce/utils/makeStructuredData_DLC.py

@@ -116,7 +116,7 @@
     # # The pred_dlc frame indices go 0:pred_dlc.shape[0]
     # #
     # # So we walk thru the dannce pred sampleIDs (we need to track these because
-    # # some could be discarded due to COM error thresholding), find its matchign frame in the matched frames,
+    # # some could be discarded due to COM error thresholding), find its matching frame in the matched frames,
     # # then associate the pred_dlc prediction with that frame
     # frames_dlc = np.arange(pred_dlc.shape[0])
 
@@ -129,7 +129,7 @@
     #         # raise Exception("Could not find sampleID in matched frames")
     #         raise Exception("Could not find sampleID in pred sampleIDs")
 
-    # Make sure we onyl take sampleIDs that are also in the DANNCE predictions
+    # Make sure we only take sampleIDs that are also in the DANNCE predictions
     dlc["sampleID"], indies, _ = np.intersect1d(
         dlc["sampleID"], pred["sampleID"], return_indices=True
     )

dannce/utils/makeSyncFiles.py

@@ -4,7 +4,7 @@
 
 Use this script if you know your camera frames are triggered and reliably synchronized.
 
-If your cameras are not natively synchronized, but you can collect timestaps for each
+If your cameras are not natively synchronized, but you can collect timestamps for each
     frame, MatchedFrames files should be generated by `preprocess_data.m`, together
     with a formatted `.mat` file listing the frameID for each camera and each timepoint.
     See `/dannce/utils/example_matchedframes.mat` file for how these timestamp data

multi-camera-calibration/calibration_extrinsic_Lframe.m

@@ -65,7 +65,7 @@
 %% Use selected points to calculate camera extrinsics
 for kk = 1:numel(lframe)
     %kk
-    % Do grid search over parms
+    % Do grid search over params
     curr_err = 1e10;
     c_save = 0;
     mr_save = 0;

tests/configs/base_config_temp.yaml

@@ -102,7 +102,7 @@ rotate: True
 # Whether to apply lens distortion during sampling. Default True
 distort: True
 
-# If true, intializes an "AVG" version of the network (i.e. final spatial expected value output layer). If false, "MAX" version
+# If true, initializes an "AVG" version of the network (i.e. final spatial expected value output layer). If false, "MAX" version
 expval: False
 
 # COM finder output confidence scores less than this threshold will be discarded