DeepLabCut · n-poulsen · Jul 19, 2024 · Apr 6, 2023 · Apr 20, 2023 · Apr 25, 2023
diff --git a/deeplabcut/modelzoo/generalized_data_converter/datasets/base.py b/deeplabcut/modelzoo/generalized_data_converter/datasets/base.py
@@ -153,20 +153,28 @@ def populate_generic(self):
         raise NotImplementedError("Must implement this function")
 
     def materialize(
-        self, proj_root, framework="coco", deepcopy=False, append_image_id=True
+        self,
+        proj_root,
+        framework="coco",
+        deepcopy=False,
+        append_image_id=True,
+        no_image_copy=False,
     ):
         mat_func = mat_func_factory(framework)
         self.meta["mat_datasets"] = {self.meta["dataset_name"]: self}
         self.meta["imageid2datasetname"] = self.imageid2datasetname
+        kwargs = dict(deepcopy=deepcopy, append_image_id=append_image_id)
+        if framework == "coco":
+            kwargs["no_image_copy"] = no_image_copy
+
         mat_func(
             proj_root,
             self.generic_train_images,
             self.generic_test_images,
             self.generic_train_annotations,
             self.generic_test_annotations,
             self.meta,
-            deepcopy=deepcopy,
-            append_image_id=append_image_id,
+            **kwargs,
         )
 
     def whether_anno_image_match(self, images, annotations):

diff --git a/deeplabcut/modelzoo/generalized_data_converter/datasets/materialize.py b/deeplabcut/modelzoo/generalized_data_converter/datasets/materialize.py
@@ -12,6 +12,7 @@
 import os
 import pickle
 import shutil
+from pathlib import Path
 
 import numpy as np
 import pandas as pd
@@ -660,9 +661,10 @@ def _generic2coco(
     train_annotations,
     test_annotations,
     meta,
-    deepcopy=False,
-    full_image_path=True,
-    append_image_id=True,
+    deepcopy: bool = False,
+    full_image_path: bool = True,
+    append_image_id: bool = True,
+    no_image_copy: bool = False,
 ):
     """
     Take generic data and create coco structure
@@ -672,6 +674,17 @@ def _generic2coco(
     annotations
     - train.json
     - test.json
+
+    Args:
+        deepcopy: Only when no_image_copy=False. If False, images are not copied from
+            their original location and symlinks are created instead.
+        full_image_path: Only when no_image_copy=False. If True, the ``file_name`` for
+            the images in the annotation files contain the resolved path to the images.
+            Otherwise, a relative path is used.
+        append_image_id: Only when no_image_copy=False. Appends the image IDs in the
+            dataset to the image names.
+        no_image_copy: Instead of copying images to the COCO dataset, the full paths to
+            the images in the original dataset are used in the annotations.
     """
 
     os.makedirs(os.path.join(proj_root, "images"), exist_ok=True)
@@ -693,54 +706,46 @@ def _generic2coco(
     broken_links = []
     # copying images via symbolic link
     for image in train_images + test_images:
-        src = image["file_name"]
+        # important to resolve the filepath! Otherwise, errors can occur when running
+        # this code from Jupyter Notebooks
+        src = Path(image["file_name"]).resolve()
         image_id = image["id"]
 
-        if not os.path.exists(src):
+        if not src.exists():
             print("problem comes from", image["source_dataset"])
             print(src)
             broken_links.append(image_id)
             continue
-        else:
-            pass
-            # print ('success comes from', image['source_dataset'])
-            # print (src)
-
-        # in dlc, some images have same name but under different folder
-        # we used to use a parent folder to distinguish them, but it's only applicable to DLC
-        # so here it's easier to just append a id into the filename
 
-        image_name = src.split(os.sep)[-1]
+        file_name = str(src)
+        dest = src
+        if not no_image_copy:
+            # in dlc, some images have same name but under different folder
+            # we used to use a parent folder to distinguish them, but it's only
+            # applicable to DLC so here it's easier to append an id into the filename
 
-        if image_name.count(".") > 1:
-            sep = image_name.rfind(".")
-            pre, suffix = image_name[:sep], image_name[sep + 1 :]
-        else:
-            # this does not work for image file that looks like image9.5.jpg..
-            pre, suffix = image_name.split(".")
-
-        # not to repeatedly add image id in memory replay training
-        if append_image_id:
-            dest_image_name = f"{pre}_{image_id}.{suffix}"
-        else:
-            dest_image_name = image_name
-        dest = os.path.join(proj_root, "images", dest_image_name)
+            # not to repeatedly add image id in memory replay training
+            dest_image_name = src.name
+            if append_image_id:
+                dest_image_name = f"{src.stem}_{image_id}{src.suffix}"
 
-        # now, we will also need to update the path in the config files
+            dest = Path(proj_root) / "images" / dest_image_name
+            dest = dest.resolve()
 
-        if full_image_path:
-            image["file_name"] = dest
-        else:
-            image["file_name"] = os.path.join("images", dest_image_name)
+            file_name = str(Path(*dest.parts[-2:]))
+            if full_image_path:
+                file_name = str(dest)
 
-        if deepcopy:
-            shutil.copy(src, dest)
-        else:
-            try:
-                os.symlink(src, dest)
-            except:
-                pass
+            if deepcopy:
+                shutil.copy(src, dest)
+            else:
+                try:
+                    os.symlink(src, dest)
+                except Exception as err:
+                    print(f"Could not create a symlink from {src} to {dest}: {err}")
+                    pass
 
+        image["file_name"] = file_name
         lookuptable[dest] = src
 
     train_annotations = [

diff --git a/deeplabcut/pose_estimation_pytorch/apis/__init__.py b/deeplabcut/pose_estimation_pytorch/apis/__init__.py
@@ -9,7 +9,10 @@
 # Licensed under GNU Lesser General Public License v3.0
 #
 
-from deeplabcut.pose_estimation_pytorch.apis.analyze_images import analyze_images
+from deeplabcut.pose_estimation_pytorch.apis.analyze_images import (
+    analyze_images,
+    superanimal_analyze_images,
+)
 from deeplabcut.pose_estimation_pytorch.apis.analyze_videos import analyze_videos
 from deeplabcut.pose_estimation_pytorch.apis.convert_detections_to_tracklets import (
     convert_detections2tracklets,

diff --git a/deeplabcut/pose_estimation_pytorch/apis/analyze_images.py b/deeplabcut/pose_estimation_pytorch/apis/analyze_images.py
@@ -47,33 +47,37 @@ def superanimal_analyze_images(
     images: str | Path | list[str] | list[Path],
     max_individuals: int,
     out_folder: str,
+    bbox_threshold: float = 0.6,
     progress_bar: bool = True,
     device: str | None = None,
     customized_pose_checkpoint: str | None = None,
     customized_detector_checkpoint: str | None = None,
     customized_model_config: str | None = None,
 ):
     """
-    This funciton inferences a superanimal model on a set of images and saves the results as labeled images.
+    This funciton inferences a superanimal model on a set of images and saves the
+    results as labeled images.
 
     Parameters
     ----------
     superanimal_name: str
-        The name of the superanimal to analyze.
-        supported list:
-        superanimal_topviewmouse
-        superanimal_quadruped
+        The name of the superanimal to analyze. Supported list:
+            - "superanimal_topviewmouse"
+            - "superanimal_quadruped"
     model_name: str
-        The name of the model to use for inference.
-        supported list:
-        hrnetw32
+        The name of the model to use for inference. Supported list:
+            - "hrnetw32"
     images: str | Path | list[str] | list[Path]
         The images to analyze. Can either be a directory containing images, or
         a list of paths of images.
     max_individuals: int
         The maximum number of individuals to detect in each image.
     out_folder: str
         The directory where the labeled images will be saved.
+    bbox_threshold: float, default=0.1
+        The minimum confidence score to keep bounding box detections. Must be in (0, 1).
+        Only used when `customized_model_config=None` (otherwise, edit your
+        `customized_model_config` with the desired bbox_threshold).
     progress_bar: bool
         Whether to display a progress bar when running inference.
     device: str | None
@@ -95,17 +99,19 @@ def superanimal_analyze_images(
     --------
     >>> import deeplabcut
     >>> from deeplabcut.pose_estimation_pytorch.apis.analyze_images import superanimal_analyze_images
-    >>> superanimal_name = 'superanimal_quadruped'
-    >>> model_name = 'hrnetw32'
-    >>> device = 'cuda'
+    >>> superanimal_name = "superanimal_quadruped"
+    >>> model_name = "hrnetw32"
+    >>> device = "cuda"
     >>> max_individuals = 3
-    >>> test_images_folder = 'test_rodent_images'
-    >>> out_images_folder = 'vis_test_rodent_images'
-    >>> ret = superanimal_analyze_images(superanimal_name,
-                                model_name,
-                                test_images_folder,
-                                max_individuals,
-                                out_images_folder)
+    >>> test_images_folder = "test_rodent_images"
+    >>> out_images_folder = "vis_test_rodent_images"
+    >>> ret = superanimal_analyze_images(
+    >>>     superanimal_name,
+    >>>     model_name,
+    >>>     test_images_folder,
+    >>>     max_individuals,
+    >>>     out_images_folder
+    >>> )
     """
 
     os.makedirs(out_folder, exist_ok=True)
@@ -119,6 +125,10 @@ def superanimal_analyze_images(
             snapshot_path,
             detector_path,
         ) = get_config_model_paths(superanimal_name, model_name)
+
+        if "detector" in model_cfg:
+            model_cfg["detector"]["model"]["box_score_thresh"] = bbox_threshold
+
         config = {**project_config, **model_cfg}
         config = update_config(config, max_individuals, device)        
     else:
@@ -146,9 +156,7 @@ def superanimal_analyze_images(
 
     superanimal_colormaps = get_superanimal_colormaps()
     colormap = superanimal_colormaps[superanimal_name]
-
     create_labeled_images_from_predictions(predictions, out_folder, colormap)
-
     return predictions
 
 
@@ -164,8 +172,6 @@ def analyze_images(
     device: str | None = None,
     max_individuals: int | None = None,
     progress_bar: bool = True,
-    superanimal_name=None,
-    model_name=None,
 ) -> dict[str, dict]:
     """Runs analysis on images using a pose model.
 

diff --git a/deeplabcut/pose_estimation_pytorch/apis/train.py b/deeplabcut/pose_estimation_pytorch/apis/train.py
@@ -257,8 +257,10 @@ def train_network(
             dataset_params = loader.get_dataset_parameters()
             backbone_name = loader.model_cfg["model"]["backbone"]["model_name"]
             model_name = modelzoo_utils.get_pose_model_type(backbone_name)
-            # at some point train_network should support a different train_file passing so memory replay can also take the same train file
+            # at some point train_network should support a different train_file passing
+            # so memory replay can also take the same train file
 
+            print("Preparing data for memory replay (this can take some time)")
             prepare_memory_replay(
                 loader.project_path,
                 shuffle,
@@ -271,6 +273,7 @@ def train_network(
                 customized_pose_checkpoint=weight_init.customized_pose_checkpoint,
             )
 
+            print("Loading memory replay data")
             loader = COCOLoader(
                 project_root=Path(loader.model_folder).parent / "memory_replay",
                 model_config_path=loader.model_config_path,

diff --git a/deeplabcut/pose_estimation_pytorch/config/base/base.yaml b/deeplabcut/pose_estimation_pytorch/config/base/base.yaml
@@ -5,7 +5,7 @@ runner:
   gpus: null
   key_metric: "test.mAP"
   key_metric_asc: true
-  eval_interval: 1
+  eval_interval: 10
   optimizer:
     type: AdamW
     params: