diff --git a/pyproject.toml b/pyproject.toml
index 966e2b00..5c7f9c30 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -23,25 +23,25 @@ dependencies = [
     "flake8-unused-arguments>=0.0.13",
     "fvcore>=0.1.5",
     "lightning>=2.2.1",
-    "matplotlib>=3.8.3",
-    "monai>=1.3.0",
     "nibabel>=5.2.1",
     "numpy>=1.26.4",
-    "pandas>=2.2.1",
     "python-dotenv==1.0.0",
     "scikit_image>=0.22.0",
     "scikit_learn>=1.4.1.post1",
-    "seaborn>=0.13.2",
-    "SimpleITK>=2.3.1",
     "tqdm>=4.66.2",
     "timm>=0.9.8",
     "torchmetrics>=1.4.0.post0",
     "wandb>=0.16.3",
-    "weave>=0.39.0",
 ]
 
 
 [project.optional-dependencies]
+extras = [
+    "matplotlib>=3.8.3",
+    "monai>=1.3.0",
+    "SimpleITK>=2.3.1",
+    "pandas>=2.2.1",
+]
 test = [
     'pytest>=7.3',
     'flake8>=6.1.0',
diff --git a/yucca/documentation/templates/functional_inference.py b/yucca/documentation/templates/functional_inference.py
index 424d1822..f00c7e1c 100644
--- a/yucca/documentation/templates/functional_inference.py
+++ b/yucca/documentation/templates/functional_inference.py
@@ -2,7 +2,6 @@
     import lightning as L
     import os
     import torch
-    from batchgenerators.utilities.file_and_folder_operations import maybe_mkdir_p as ensure_dir_exists
     from yucca.paths import (
         get_models_path,
         get_results_path,
@@ -41,7 +40,7 @@
         "version_0",
         "best",
     )
-    ensure_dir_exists(save_path)
+    os.makedirs(save_path, exist_ok=True)
 
     ckpt = torch.load(ckpt_path, map_location="cpu")
     pred_writer = WritePredictionFromLogits(output_dir=save_path, save_softmax=False, write_interval="batch")
diff --git a/yucca/documentation/templates/functional_preprocessing.py b/yucca/documentation/templates/functional_preprocessing.py
index 64d78ced..5333ce30 100644
--- a/yucca/documentation/templates/functional_preprocessing.py
+++ b/yucca/documentation/templates/functional_preprocessing.py
@@ -3,26 +3,22 @@
     import os
     import numpy as np
     import torch
-    from batchgenerators.utilities.file_and_folder_operations import (
-        subfiles,
-        join,
-        save_pickle,
-        maybe_mkdir_p as ensure_dir_exists,
-    )
+    from yucca.functional.utils.files_and_folders import subfiles
+    from yucca.functional.utils.saving import save_pickle
     from yucca.paths import get_raw_data_path, get_preprocessed_data_path
     from yucca.documentation.templates.template_config import config
     from yucca.functional.preprocessing import preprocess_case_for_training_with_label, preprocess_case_for_inference
     from yucca.functional.utils.loading import read_file_to_nifti_or_np
 
-    raw_images_dir = join(get_raw_data_path(), config["task"], "imagesTr")
-    raw_labels_dir = join(get_raw_data_path(), config["task"], "labelsTr")
-    test_raw_images_dir = join(get_raw_data_path(), config["task"], "imagesTs")
+    raw_images_dir = os.path.join(get_raw_data_path(), config["task"], "imagesTr")
+    raw_labels_dir = os.path.join(get_raw_data_path(), config["task"], "labelsTr")
+    test_raw_images_dir = os.path.join(get_raw_data_path(), config["task"], "imagesTs")
 
-    target_dir = join(get_preprocessed_data_path(), config["task"], config["config_name"])
-    test_target_dir = join(get_preprocessed_data_path(), config["task"] + "_test", config["config_name"])
+    target_dir = os.path.join(get_preprocessed_data_path(), config["task"], config["config_name"])
+    test_target_dir = os.path.join(get_preprocessed_data_path(), config["task"] + "_test", config["config_name"])
 
-    ensure_dir_exists(target_dir)
-    ensure_dir_exists(test_target_dir)
+    os.makedirs(target_dir, exist_ok=True)
+    os.makedirs(test_target_dir, exist_ok=True)
 
     # Preprocess the training data
     subjects = [file[: -len(config["extension"])] for file in subfiles(raw_labels_dir, join=False) if not file.startswith(".")]
@@ -34,7 +30,7 @@
             if re.search(re.escape(sub) + "_" + r"\d{3}" + ".", os.path.split(image_path)[-1])
         ]
         images = [read_file_to_nifti_or_np(image) for image in images]
-        label = read_file_to_nifti_or_np(join(raw_labels_dir, sub + config["extension"]))
+        label = read_file_to_nifti_or_np(os.path.join(raw_labels_dir, sub + config["extension"]))
         images, label, image_props = preprocess_case_for_training_with_label(
             images=images,
             label=label,
@@ -45,7 +41,7 @@
         )
         images = np.vstack((np.array(images), np.array(label)[np.newaxis]), dtype=np.float32)
 
-        save_path = join(target_dir, sub)
+        save_path = os.path.join(target_dir, sub)
         np.save(save_path + ".npy", images)
         save_pickle(image_props, save_path + ".pkl")
 
@@ -73,6 +69,6 @@
             target_spacing=config["target_spacing"],
             target_orientation=config["target_coordinate_system"],
         )
-        save_path = join(test_target_dir, sub)
+        save_path = os.path.join(test_target_dir, sub)
         torch.save(images, save_path + ".pt")
         save_pickle(image_props, save_path + ".pkl")
diff --git a/yucca/documentation/tests/transforms/dataloader.ipynb b/yucca/documentation/tests/transforms/dataloader.ipynb
index 41ac220d..391b6c19 100644
--- a/yucca/documentation/tests/transforms/dataloader.ipynb
+++ b/yucca/documentation/tests/transforms/dataloader.ipynb
@@ -2,19 +2,20 @@
  "cells": [
   {
    "cell_type": "code",
-   "execution_count": 1,
+   "execution_count": null,
    "metadata": {},
    "outputs": [],
    "source": [
     "from yucca.modules.data.datasets.YuccaDataset import YuccaTrainDataset\n",
     "from yucca.paths import yucca_preprocessed_data\n",
-    "from batchgenerators.utilities.file_and_folder_operations import join, subfiles\n",
+    "from yucca.functional.utils.files_and_folders import subfiles\n",
+    "import os\n",
     "from torch.utils.data import DataLoader"
    ]
   },
   {
    "cell_type": "code",
-   "execution_count": 2,
+   "execution_count": null,
    "metadata": {},
    "outputs": [
     {
@@ -28,7 +29,7 @@
    "source": [
     "from matplotlib import pyplot as plt\n",
     "\n",
-    "samples = subfiles(join(yucca_preprocessed_data(), \"Task299_Combine\", \"UnsupervisedPlanner\"), suffix=\".npy\")\n",
+    "samples = subfiles(os.path.join(yucca_preprocessed_data(), \"Task299_Combine\", \"UnsupervisedPlanner\"), suffix=\".npy\")\n",
     "\n",
     "\n",
     "dataset = YuccaTrainDataset(samples=samples, patch_size=(96,) * 3, composed_transforms=None, task_type=\"contrastive\")\n",
@@ -93,7 +94,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "samples = subfiles(join(yucca_preprocessed_data(), \"Task001_OASIS\", \"YuccaPlanner\"), suffix=\".npy\")\n",
+    "samples = subfiles(os.path.join(yucca_preprocessed_data(), \"Task001_OASIS\", \"YuccaPlanner\"), suffix=\".npy\")\n",
     "\n",
     "print(samples)\n",
     "\n",
diff --git a/yucca/functional/evaluation/__init__.py b/yucca/functional/evaluation/__init__.py
index 1203c213..ebffa940 100644
--- a/yucca/functional/evaluation/__init__.py
+++ b/yucca/functional/evaluation/__init__.py
@@ -1,4 +1,3 @@
-from .confusion_matrix import torch_confusion_matrix_from_logits, torch_get_tp_fp_tn_fn
 from .metrics import (
     dice,
     dice_per_label,
@@ -17,5 +16,4 @@
     total_pos_gt,
     total_pos_pred,
 )
-from .obj_metrics import get_obj_stats_for_label, obj_get_tp_fp_fn_gtvols_predvols
 from .surface_metrics import get_surface_metrics_for_label
diff --git a/yucca/functional/evaluation/evaluate_folder.py b/yucca/functional/evaluation/evaluate_folder.py
index 9dd989b4..1a04b222 100644
--- a/yucca/functional/evaluation/evaluate_folder.py
+++ b/yucca/functional/evaluation/evaluate_folder.py
@@ -2,6 +2,7 @@
 import numpy as np
 import nibabel as nib
 import logging
+import os
 from typing import Optional
 from yucca.functional.transforms.label_transforms import convert_labels_to_regions, translate_region_labels
 from yucca.functional.utils.nib_utils import get_nib_spacing
@@ -9,7 +10,6 @@
 from yucca.functional.evaluation.obj_metrics import get_obj_stats_for_label
 from yucca.functional.evaluation.surface_metrics import get_surface_metrics_for_label
 from tqdm import tqdm
-from batchgenerators.utilities.file_and_folder_operations import join
 from sklearn.metrics import confusion_matrix
 from yucca.functional.evaluation.metrics import auroc
 
@@ -97,8 +97,8 @@ def evaluate_multilabel_case_segm(
     assert regions is not None
 
     case_dict = {}
-    predpath = join(folder_with_predictions, case)
-    gtpath = join(folder_with_ground_truth, case)
+    predpath = os.path.join(folder_with_predictions, case)
+    gtpath = os.path.join(folder_with_ground_truth, case)
     case_dict["prediction_path"] = predpath
     case_dict["ground_truth_path"] = gtpath
 
@@ -184,8 +184,8 @@ def evaluate_case_segm(
     surface_tol: int = 1,
 ):
     case_dict = {}
-    predpath = join(folder_with_predictions, case)
-    gtpath = join(folder_with_ground_truth, case)
+    predpath = os.path.join(folder_with_predictions, case)
+    gtpath = os.path.join(folder_with_ground_truth, case)
 
     case_dict["prediction_path"] = predpath
     case_dict["ground_truth_path"] = gtpath
@@ -269,8 +269,8 @@ def evaluate_folder_cls(
 
     # load predictions and ground truths
     for case in tqdm(subjects, desc="Evaluating"):
-        predpath = join(folder_with_predictions, case)
-        gtpath = join(folder_with_ground_truth, case)
+        predpath = os.path.join(folder_with_predictions, case)
+        gtpath = os.path.join(folder_with_ground_truth, case)
 
         pred: int = np.loadtxt(predpath)
         gt: int = np.loadtxt(gtpath)
diff --git a/yucca/functional/planning.py b/yucca/functional/planning.py
index 4b280a74..28b0f479 100644
--- a/yucca/functional/planning.py
+++ b/yucca/functional/planning.py
@@ -1,6 +1,7 @@
 import numpy as np
 from typing import Optional, List, Union
-from batchgenerators.utilities.file_and_folder_operations import subfiles, load_pickle
+from yucca.functional.utils.files_and_folders import subfiles
+from yucca.functional.utils.loading import load_pickle
 
 
 def make_plans_file(
diff --git a/yucca/functional/utils/files_and_folders.py b/yucca/functional/utils/files_and_folders.py
index 276e8e51..d54f1bfa 100644
--- a/yucca/functional/utils/files_and_folders.py
+++ b/yucca/functional/utils/files_and_folders.py
@@ -4,11 +4,29 @@
 import re
 import shutil
 import os
-from batchgenerators.utilities.file_and_folder_operations import (
-    join,
-    subdirs,
-)
-from typing import Union, List
+from typing import Union, List, Optional
+
+
+def subdirs(
+    folder: str, join: bool = True, prefix: Optional[str] = None, suffix: Optional[str] = None, sort: bool = True
+) -> List[str]:
+    """
+    implementation by: https://github.com/MIC-DKFZ/batchgenerators
+    """
+    subdirectories = []
+    with os.scandir(folder) as entries:
+        for entry in entries:
+            if (
+                entry.is_dir()
+                and (prefix is None or entry.name.startswith(prefix))
+                and (suffix is None or entry.name.endswith(suffix))
+            ):
+                dir_path = entry.path if join else entry.name
+                subdirectories.append(dir_path)
+
+    if sort:
+        subdirectories.sort()
+    return subdirectories
 
 
 def replace_in_file(file_path, pattern_replacement):
@@ -139,7 +157,7 @@ def _recursive_find_python_class(folder: list, class_name: str, current_module:
                 if ispkg:
                     next_current_module = current_module + "." + modname
                     tr = _recursive_find_python_class(
-                        [join(folder[0], modname)],
+                        [os.path.join(folder[0], modname)],
                         class_name,
                         current_module=next_current_module,
                     )
diff --git a/yucca/functional/utils/loading.py b/yucca/functional/utils/loading.py
index f65de112..c2ea42d0 100644
--- a/yucca/functional/utils/loading.py
+++ b/yucca/functional/utils/loading.py
@@ -2,6 +2,7 @@
 import os
 import nibabel as nib
 import numpy as np
+import pickle
 from PIL import Image
 
 
@@ -21,3 +22,14 @@ def read_file_to_nifti_or_np(imagepath, dtype=np.float32):
         return np.atleast_1d(np.genfromtxt(imagepath, delimiter=",", dtype=dtype))
     else:
         raise TypeError(f"File type invalid. Found extension: {ext} and expected one in [nii, nii.gz, png, csv, txt]")
+
+
+def load_pickle(file: str, mode: str = "rb"):
+    with open(file, mode) as f:
+        a = pickle.load(f)
+    return a
+
+
+def load_json(p):
+    with open(p, "r") as f:
+        return json.load(f)
diff --git a/yucca/functional/utils/saving.py b/yucca/functional/utils/saving.py
index 7360b59f..5b8596d3 100644
--- a/yucca/functional/utils/saving.py
+++ b/yucca/functional/utils/saving.py
@@ -1,13 +1,16 @@
 import nibabel as nib
 import numpy as np
+import os
+import pickle
 from yucca.functional.utils.softmax import softmax
 from yucca.functional.utils.nib_utils import reorient_nib_image
+from yucca.functional.utils.file_and_folders import subfiles
 from PIL import Image
-from batchgenerators.utilities.file_and_folder_operations import (
-    join,
-    subfiles,
-    maybe_mkdir_p as ensure_dir_exists,
-)
+
+
+def save_pickle(obj, file: str, mode: str = "wb") -> None:
+    with open(file, mode) as f:
+        pickle.dump(obj, f)
 
 
 def save_nifti_from_numpy(pred, outpath, properties, compression=9):
@@ -87,7 +90,7 @@ def save_multilabel_prediction_from_logits(logits, outpath, properties, compress
 
 
 def merge_softmax_from_folders(folders: list, outpath, method="sum"):
-    ensure_dir_exists(outpath)
+    os.makedirs(outpath, exists_ok=True)
     cases = subfiles(folders[0], suffix=".npz", join=False)
     for folder in folders:
         assert cases == subfiles(folder, suffix=".npz", join=False), (
@@ -98,7 +101,7 @@ def merge_softmax_from_folders(folders: list, outpath, method="sum"):
         )
 
     for case in cases:
-        files_for_case = [np.load(join(folder, case), allow_pickle=True) for folder in folders]
+        files_for_case = [np.load(os.path.join(folder, case), allow_pickle=True) for folder in folders]
         properties_for_case = files_for_case[0]["properties"]
         files_for_case = [file["data"].astype(np.float32) for file in files_for_case]
 
@@ -108,7 +111,7 @@ def merge_softmax_from_folders(folders: list, outpath, method="sum"):
         files_for_case = np.argmax(files_for_case, 0)
         save_nifti_from_numpy(
             files_for_case,
-            join(outpath, case[:-4]),
+            os.path.join(outpath, case[:-4]),
             properties=properties_for_case.item(),
         )
 
diff --git a/yucca/modules/callbacks/loggers.py b/yucca/modules/callbacks/loggers.py
index 3e52f323..346ec959 100644
--- a/yucca/modules/callbacks/loggers.py
+++ b/yucca/modules/callbacks/loggers.py
@@ -3,15 +3,10 @@
 import logging
 from argparse import Namespace
 from lightning.pytorch.loggers.logger import Logger
-from pytorch_lightning.utilities.rank_zero import rank_zero_only
-from pytorch_lightning.core.saving import save_hparams_to_yaml
-from lightning_fabric.utilities.logger import _convert_params
+from lightning.pytorch.utilities.rank_zero import rank_zero_only
+from lightning.pytorch.core.saving import save_hparams_to_yaml
+from lightning.fabric.utilities.logger import _convert_params
 from time import localtime, strftime, time
-from batchgenerators.utilities.file_and_folder_operations import (
-    join,
-    maybe_mkdir_p as ensure_dir_exists,
-    isdir,
-)
 from typing import Any, Dict, Optional, Union
 
 
@@ -60,18 +55,18 @@ def root_dir(self):
     def log_dir(self):
         log_dir = self.root_dir
         if self.name is not None:
-            log_dir = join(log_dir, self.name)
+            log_dir = os.path.join(log_dir, self.name)
         if self.version is not None:
             version = self.version if isinstance(self.version, str) else f"version_{self.version}"
-            log_dir = join(log_dir, version)
-        if not isdir(log_dir):
-            ensure_dir_exists(log_dir)
+            log_dir = os.path.join(log_dir, version)
+        if not os.path.isdir(log_dir):
+            os.makedirs(log_dir, exist_ok=True)
         return log_dir
 
     @rank_zero_only
     def create_logfile(self):
-        ensure_dir_exists(self.log_dir)
-        self.log_file = join(
+        os.makedirs(self.log_dir, exist_ok=True)
+        self.log_file = os.path.join(
             self.log_dir,
             "training_log.txt",
         )
diff --git a/yucca/modules/callbacks/prediction_writer.py b/yucca/modules/callbacks/prediction_writer.py
index 74445f1d..6427e7a0 100644
--- a/yucca/modules/callbacks/prediction_writer.py
+++ b/yucca/modules/callbacks/prediction_writer.py
@@ -1,6 +1,6 @@
 from lightning.pytorch.callbacks import BasePredictionWriter
 from yucca.functional.utils.saving import save_prediction_from_logits, save_multilabel_prediction_from_logits
-from batchgenerators.utilities.file_and_folder_operations import join
+import os
 
 
 class WritePredictionFromLogits(BasePredictionWriter):
@@ -21,13 +21,13 @@ def write_on_batch_end(self, _trainer, _pl_module, data_dict, _batch_indices, _b
         if self.multilabel:
             save_multilabel_prediction_from_logits(
                 logits,
-                join(self.output_dir, case_id),
+                os.path.join(self.output_dir, case_id),
                 properties=properties,
             )
         else:
             save_prediction_from_logits(
                 logits,
-                join(self.output_dir, case_id),
+                os.path.join(self.output_dir, case_id),
                 properties=properties,
                 save_softmax=self.save_softmax,
             )
diff --git a/yucca/modules/data/augmentation/transforms/YuccaTransform.py b/yucca/modules/data/augmentation/transforms/YuccaTransform.py
index be0ce6e2..845fba79 100644
--- a/yucca/modules/data/augmentation/transforms/YuccaTransform.py
+++ b/yucca/modules/data/augmentation/transforms/YuccaTransform.py
@@ -1,9 +1,22 @@
-from batchgenerators.transforms.abstract_transforms import AbstractTransform
-from abc import abstractmethod
+import abc
+
+
+class AbstractTransform(object):
+    __metaclass__ = abc.ABCMeta
+
+    @abc.abstractmethod
+    def __call__(self, **data_dict):
+        raise NotImplementedError("Abstract, so implement")
+
+    def __repr__(self):
+        ret_str = (
+            str(type(self).__name__) + "( " + ", ".join([key + " = " + repr(val) for key, val in self.__dict__.items()]) + " )"
+        )
+        return ret_str
 
 
 class YuccaTransform(AbstractTransform):
-    @abstractmethod
+    @abc.abstractmethod
     def get_params(self):
         """
         This will return a random value between
@@ -35,7 +48,7 @@ def get_params(self):
         data modalities and labels remain registered.
         """
 
-    @abstractmethod
+    @abc.abstractmethod
     def __call__(self):
         """
         This will be of the form __call__(self, packed_dict: dict = None, **unpacked_dict):
diff --git a/yucca/modules/data/data_modules/YuccaDataModule.py b/yucca/modules/data/data_modules/YuccaDataModule.py
index 86fb2969..ccbf2989 100644
--- a/yucca/modules/data/data_modules/YuccaDataModule.py
+++ b/yucca/modules/data/data_modules/YuccaDataModule.py
@@ -2,9 +2,9 @@
 import torchvision
 import logging
 import torch
+import os
 from typing import Literal, Optional, Union
 from torch.utils.data import DataLoader, Sampler
-from batchgenerators.utilities.file_and_folder_operations import join
 from yucca.pipeline.configuration.split_data import SplitConfig
 from yucca.modules.data.datasets.YuccaDataset import YuccaTestDataset, YuccaTrainDataset
 from yucca.modules.data.samplers import InfiniteRandomSampler
@@ -116,8 +116,8 @@ def setup(self, stage: Literal["fit", "test", "predict"]):
             assert self.splits_config is not None
             assert self.task_type is not None
 
-            self.train_samples = [join(self.train_data_dir, i) for i in self.splits_config.train(self.split_idx)]
-            self.val_samples = [join(self.train_data_dir, i) for i in self.splits_config.val(self.split_idx)]
+            self.train_samples = [os.path.join(self.train_data_dir, i) for i in self.splits_config.train(self.split_idx)]
+            self.val_samples = [os.path.join(self.train_data_dir, i) for i in self.splits_config.val(self.split_idx)]
 
             if len(self.train_samples) < 100:
                 logging.info(f"Training on samples: {self.train_samples}")
diff --git a/yucca/modules/data/datasets/YuccaCompressedDataset.py b/yucca/modules/data/datasets/YuccaCompressedDataset.py
index 760636f5..ce4c1ec4 100644
--- a/yucca/modules/data/datasets/YuccaCompressedDataset.py
+++ b/yucca/modules/data/datasets/YuccaCompressedDataset.py
@@ -1,12 +1,12 @@
 import numpy as np
-from batchgenerators.utilities.file_and_folder_operations import isfile
+import os
 from yucca.modules.data.datasets.YuccaDataset import YuccaTrainDataset
 
 
 class YuccaCompressedTrainDataset(YuccaTrainDataset):
     def load_and_maybe_keep_volume(self, path: str):
         path = path + ".npz"
-        if isfile(path):
+        if os.path.isfile(path):
             try:
                 return np.load(path, "r")["data"]
             except ValueError:
diff --git a/yucca/modules/data/datasets/YuccaDataset.py b/yucca/modules/data/datasets/YuccaDataset.py
index 54ffcf49..e52f0dec 100644
--- a/yucca/modules/data/datasets/YuccaDataset.py
+++ b/yucca/modules/data/datasets/YuccaDataset.py
@@ -3,7 +3,8 @@
 import os
 import logging
 from typing import Union, Literal, Optional
-from batchgenerators.utilities.file_and_folder_operations import subfiles, load_pickle, isfile
+from yucca.functional.utils.files_and_folders import subfiles
+from yucca.functional.utils.loading import load_pickle
 from yucca.modules.data.augmentation.transforms.cropping_and_padding import CropPad
 from yucca.modules.data.augmentation.transforms.formatting import NumpyToTorch
 
@@ -64,7 +65,7 @@ def load_and_maybe_keep_pickle(self, path):
     def load_and_maybe_keep_volume(self, path):
         path = path + ".npy"
         if not self.keep_in_ram:
-            if isfile(path):
+            if os.path.isfile(path):
                 try:
                     return np.load(path, "r")
                 except ValueError:
@@ -72,7 +73,7 @@ def load_and_maybe_keep_volume(self, path):
             else:
                 print("uncompressed data was not found.")
 
-        if isfile(path):
+        if os.path.isfile(path):
             if path in self.already_loaded_cases:
                 return self.already_loaded_cases[path]
             try:
@@ -242,10 +243,9 @@ def __getitem__(self, idx):
 if __name__ == "__main__":
     import torch
     from yucca.paths import get_preprocessed_data_path
-    from batchgenerators.utilities.file_and_folder_operations import join
     from yucca.modules.data.samplers import InfiniteRandomSampler
 
-    files = subfiles(join(get_preprocessed_data_path(), "Task001_OASIS/YuccaPlanner"), suffix="npy")
+    files = subfiles(os.path.join(get_preprocessed_data_path(), "Task001_OASIS/YuccaPlanner"), suffix="npy")
     ds = YuccaTrainDataset(files, patch_size=(12, 12, 12))
     sampler = InfiniteRandomSampler(ds)
     dl = torch.utils.data.DataLoader(ds, num_workers=2, batch_size=2, sampler=sampler)
diff --git a/yucca/modules/networks/networks/3D_resnet.py b/yucca/modules/networks/networks/3D_resnet.py
index 5ec788e3..795601f3 100644
--- a/yucca/modules/networks/networks/3D_resnet.py
+++ b/yucca/modules/networks/networks/3D_resnet.py
@@ -1,5 +1,4 @@
-from typing import Union, List, Optional, Callable, Type
-from pytorchvideo.models.resnet import create_resnet
+from typing import List, Optional, Type
 from torch import nn, Tensor
 import torch
 from yucca.modules.networks.blocks_and_layers.res_blocks import BasicBlock, Bottleneck, conv_k1
diff --git a/yucca/modules/networks/utils/model_memory_estimation.py b/yucca/modules/networks/utils/model_memory_estimation.py
index 9cd93c56..acd9a8bc 100644
--- a/yucca/modules/networks/utils/model_memory_estimation.py
+++ b/yucca/modules/networks/utils/model_memory_estimation.py
@@ -26,11 +26,10 @@
 import yucca
 import math
 import logging
+import os
 from yucca.functional.utils.torch_utils import get_available_device, flush_and_get_torch_memory_allocated
 from yucca.functional.utils.files_and_folders import recursive_find_python_class
 from yucca.functional.utils.kwargs import filter_kwargs
-
-from batchgenerators.utilities.file_and_folder_operations import join
 from torch import nn
 
 
@@ -154,7 +153,7 @@ def find_optimal_tensor_dims(
         model_name = model_name.split("_")[0]
 
     model = recursive_find_python_class(
-        folder=[join(yucca.__path__[0], "modules", "networks")],
+        folder=[os.path.join(yucca.__path__[0], "modules", "networks")],
         class_name=model_name,
         current_module="yucca.modules.networks",
     )
diff --git a/yucca/modules/optimization/utils/LossTree.py b/yucca/modules/optimization/utils/LossTree.py
index 0fd0dc22..3497a17a 100644
--- a/yucca/modules/optimization/utils/LossTree.py
+++ b/yucca/modules/optimization/utils/LossTree.py
@@ -1,4 +1,4 @@
-from batchgenerators.utilities.file_and_folder_operations import load_json
+from yucca.functional.utils.loading import load_json
 
 
 class node:
diff --git a/yucca/paths.py b/yucca/paths.py
index a6ba66c0..32ca007d 100644
--- a/yucca/paths.py
+++ b/yucca/paths.py
@@ -5,8 +5,6 @@
 import os
 from dotenv import load_dotenv
 
-from batchgenerators.utilities.file_and_folder_operations import maybe_mkdir_p as ensure_dir_exists
-
 
 def var_is_set(var):
     return var in os.environ.keys()
@@ -18,7 +16,7 @@ def get_environment_variable(var):
         raise ValueError(f"Missing required environment variable {var}.")
 
     path = os.environ[var]
-    ensure_dir_exists(path)
+    os.makedirs(path, exist_ok=True)
     return path
 
 
diff --git a/yucca/pipeline/configuration/configure_checkpoint.py b/yucca/pipeline/configuration/configure_checkpoint.py
index d1b9dbe5..a27e9e93 100644
--- a/yucca/pipeline/configuration/configure_checkpoint.py
+++ b/yucca/pipeline/configuration/configure_checkpoint.py
@@ -1,5 +1,5 @@
+import os
 import torch
-from batchgenerators.utilities.file_and_folder_operations import join, isfile
 from dataclasses import dataclass
 from typing import Union
 from yucca.pipeline.configuration.configure_paths import PathConfig
@@ -96,12 +96,16 @@ def get_checkpoint_config_from_ckpt(ckpt_path: str):
 
 def find_checkpoint_path(ckpt_path: Union[str, None], continue_from_most_recent: bool, version: int, version_dir: str):
     if ckpt_path:
-        assert isfile(ckpt_path), f"Checkpoint was not found. Looked in: {ckpt_path}"
+        assert os.path.isfile(ckpt_path), f"Checkpoint was not found. Looked in: {ckpt_path}"
         logging.info(f"Using ckpt file: {ckpt_path}")
         return ckpt_path
-    elif version is not None and continue_from_most_recent and isfile(join(version_dir, "checkpoints", "last.ckpt")):
+    elif (
+        version is not None
+        and continue_from_most_recent
+        and os.path.isfile(os.path.join(version_dir, "checkpoints", "last.ckpt"))
+    ):
         logging.info("Using last checkpoint and continuing training")
-        return join(version_dir, "checkpoints", "last.ckpt")
+        return os.path.join(version_dir, "checkpoints", "last.ckpt")
     else:
         return None
 
diff --git a/yucca/pipeline/configuration/configure_paths.py b/yucca/pipeline/configuration/configure_paths.py
index f70677d6..6b8cd37b 100644
--- a/yucca/pipeline/configuration/configure_paths.py
+++ b/yucca/pipeline/configuration/configure_paths.py
@@ -1,4 +1,5 @@
-from batchgenerators.utilities.file_and_folder_operations import join, isdir, subdirs, maybe_mkdir_p as ensure_dir_exists
+import os
+from yucca.functional.utils.files_and_folders import subdirs
 from dataclasses import dataclass
 from typing import Union, Literal
 from yucca.paths import get_models_path, get_preprocessed_data_path
@@ -25,9 +26,9 @@ def lm_hparams(self):
 
 
 def get_path_config(task_config: TaskConfig, stage: Literal["fit", "test", "predict"]):
-    task_dir = join(get_preprocessed_data_path(), task_config.task)
-    train_data_dir = join(task_dir, task_config.planner_name)
-    save_dir = join(
+    task_dir = os.path.join(get_preprocessed_data_path(), task_config.task)
+    train_data_dir = os.path.join(task_dir, task_config.planner_name)
+    save_dir = os.path.join(
         get_models_path(),
         task_config.task,
         task_config.model_name + "__" + task_config.model_dimensions,
@@ -37,16 +38,16 @@ def get_path_config(task_config: TaskConfig, stage: Literal["fit", "test", "pred
     )
 
     version = detect_version(save_dir, task_config.continue_from_most_recent)
-    version_dir = join(save_dir, f"version_{version}")
-    ensure_dir_exists(version_dir)
+    version_dir = os.path.join(save_dir, f"version_{version}")
+    os.makedirs(version_dir, exist_ok=True)
 
     # First try to load torch checkpoints and extract plans and carry-over information from there.
     if stage == "fit":
-        plans_path = join(task_dir, task_config.planner_name, task_config.planner_name + "_plans.json")
+        plans_path = os.path.join(task_dir, task_config.planner_name, task_config.planner_name + "_plans.json")
     if stage == "test":
         raise NotImplementedError
     if stage == "predict":
-        plans_path = join(version_dir, "hparams.yaml")
+        plans_path = os.path.join(version_dir, "hparams.yaml")
 
     return PathConfig(
         plans_path=plans_path,
@@ -60,7 +61,7 @@ def get_path_config(task_config: TaskConfig, stage: Literal["fit", "test", "pred
 
 def detect_version(save_dir, continue_from_most_recent) -> Union[None, int]:
     # If the dir doesn't exist we return version 0
-    if not isdir(save_dir):
+    if not os.path.isdir(save_dir):
         return 0
 
     # The dir exists. Check if any previous version exists in dir.
diff --git a/yucca/pipeline/configuration/configure_plans.py b/yucca/pipeline/configuration/configure_plans.py
index a3e7277b..bdca99c5 100644
--- a/yucca/pipeline/configuration/configure_plans.py
+++ b/yucca/pipeline/configuration/configure_plans.py
@@ -1,12 +1,12 @@
 import yucca
-from batchgenerators.utilities.file_and_folder_operations import join, load_json, isfile
+import os
 from dataclasses import dataclass
 from typing import Union, Literal
 from yucca.pipeline.preprocessing.UnsupervisedPreprocessor import UnsupervisedPreprocessor
 from yucca.pipeline.preprocessing.ClassificationPreprocessor import ClassificationPreprocessor
 from yucca.functional.utils.dict import without_keys
 from yucca.functional.utils.files_and_folders import recursive_find_python_class
-from yucca.functional.utils.loading import load_yaml
+from yucca.functional.utils.loading import load_yaml, load_json
 import logging
 
 
@@ -54,7 +54,7 @@ def get_plan_config(
     if stage == "predict":
         # In this case we don't want to rely on plans being found in the preprocessed folder,
         # as it might not exist.
-        if isfile(plans_path):
+        if os.path.isfile(plans_path):
             plans = load_yaml(plans_path)["config"]["plans"]
         else:
             assert ckpt_plans is not None
@@ -109,7 +109,7 @@ def setup_task_type(plans):
 
     # If key is not present in plan then we try to infer the task_type from the Type of Preprocessor
     preprocessor_class = recursive_find_python_class(
-        folder=[join(yucca.__path__[0], "pipeline", "preprocessing")],
+        folder=[os.path.join(yucca.__path__[0], "pipeline", "preprocessing")],
         class_name=plans["preprocessor"],
         current_module="yucca.pipeline.preprocessing",
     )
diff --git a/yucca/pipeline/evaluation/YuccaEvaluator.py b/yucca/pipeline/evaluation/YuccaEvaluator.py
index 375fcee6..956d556f 100644
--- a/yucca/pipeline/evaluation/YuccaEvaluator.py
+++ b/yucca/pipeline/evaluation/YuccaEvaluator.py
@@ -231,37 +231,3 @@ def save_as_json(self, dict):
         print("Saving results.json \n \n ########################################################################")
         with open(self.outpath, "w") as f:
             json.dump(dict, f, default=float, indent=4)
-
-    def update_streamtable(self, results_dict):
-        """
-        Save evaluation results to a wandb StreamTable
-
-        :param results_dict: dictionary with evaluation results
-        """
-        from weave.monitoring import StreamTable
-
-        task = self.outpath.split(os.path.sep)[-5]
-        target = self.outpath.split(os.path.sep)[-6]
-        model_name = "/".join(self.outpath.split(os.path.sep)[-4:])
-
-        st = StreamTable(table_name=task, entity_name=wandb.api.viewer()["entity"], project_name="Yucca")
-
-        stream_dict = {"0. Experiment": model_name, "0. Target Task": target}
-
-        if self.task_type == "classification":
-            stream_dict = {**stream_dict, **results_dict}
-
-        elif self.task_type == "segmentation":
-            for key, _ in results_dict.items():
-                if key == "0":
-                    continue
-                else:
-                    stream_dict.update(
-                        {f"{key}. " + k: v for k, v in results_dict[key].items() if k in self.metrics_included_in_streamtable}
-                    )
-
-        else:
-            raise NotImplementedError("Task type not supported")
-
-        st.log(stream_dict)
-        st.finish()