Sami's Developpement

bash/deploy_win64.sh

@@ -0,0 +1,7 @@
+
+# create a venv :
+python -m venv NAME_OF_YOUR_VIRTUAL_ENV
+# then install :
+pip install pyinstaller python-tk numpy SimpleITK scikit-image tqdm paramiko pyyaml matplotlib 
+# run the pyinstaller (with console) and exclude modules
+pyinstaller -F --clean --upx-dir PATH_TO_UPX   PATH_TO_GUI\gui.py --exclude=biom3d --exclude=torch --hidden-import='PIL._tkinter_finder'

docs/quick_run_gui.md

@@ -18,7 +18,11 @@ Biom3d comes with 2 modes: local or remote. 'Local' means that the computation w
 
 If you have installed biom3d with the local version simply click on the 'Start locally' button to start, you can choose a path to store your files in the field over the button, by default, the files are stored in the directory where biom3d have been launched.
 
+<<<<<<< HEAD
+If you have installed biom3d with the remote version, you must then complete the required fields. The first one is the IP address of your remote computer (where the API of biom3d is installed). The second and third one is your user name and password to connect to the remote computer, the forth one is the path to your virtual environment (if you don't have a virtual environment leave it empty ).
+=======
 If you have installed biom3d with the remote version, you must then complete the required fields. The first one is the IP address of your remote computer (where the API of biom3d is installed). The second and third one is your user name and password to connect to the remote computer, the forth one is the path to your virtual environment (if you don't have a virtual environment leave it empty).
+>>>>>>> f562750641a74b960ed698e427ed811313552d0e
 
 
 ## Preprocess & Train

src/biom3d/auto_config.py

@@ -38,6 +38,10 @@ def compute_median(path, return_spacing=False):
     for i in range(len(path_imgs)):
 
         img,metadata = adaptive_imread(path_imgs[i])
+        # Check if the image is 2D (has two dimensions)
+        if len(img.shape) == 2:
+            # Add a third dimension with size 1 to make it 3D
+            img = np.expand_dims(img, axis=0)
         spacing = None if not 'spacing' in metadata.keys() else metadata['spacing']
 
         assert len(img.shape)>0, "[Error] Wrong image image."
@@ -91,6 +95,10 @@ def data_fingerprint(img_dir, msk_dir=None, num_samples=10000):
 
     for i in range(len(path_imgs)):
         img,metadata = adaptive_imread(path_imgs[i])
+        # Check if the image is 2D (has two dimensions)
+        if len(img.shape) == 2:
+            # Add a third dimension with size 1 to make it 3D
+            img = np.expand_dims(img, axis=0)
         spacing = None if not 'spacing' in metadata.keys() else metadata['spacing']
 
         # store the size
@@ -103,6 +111,10 @@ def data_fingerprint(img_dir, msk_dir=None, num_samples=10000):
         if msk_dir is not None:
             # read msk
             msk,_ = adaptive_imread(path_msks[i])
+            # Check if the image is 2D (has two dimensions)
+            if len(msk.shape) == 2:
+                # Add a third dimension with size 1 to make it 3D
+                msk = np.expand_dims(msk, axis=0)
 
             # extract only useful voxels
             img = img[msk > 0]
@@ -151,7 +163,7 @@ def find_patch_pool_batch(dims, max_dims=(128,128,128), max_pool=5, epsilon=1e-3
     batch: numpy.ndarray
         Batch size.
     """
-    # transform tuples into arrays
+    # transform tuples into arrays      
     assert len(dims)==3 or len(dims)==4, print("Dims has not the correct number of dimensions: len(dims)=", len(dims))
     if len(dims)==4:
         dims=dims[1:]
@@ -279,15 +291,23 @@ def get_aug_patch(patch_size):
 
 # ----------------------------------------------------------------------------
 # Display 
+def parameters_return(patch, pool, batch, config_path):
+    print(batch)
+    print(patch)
+    aug_patch= get_aug_patch(patch)
+    print(aug_patch)
+    print(pool)
+    print(config_path)
 
 def display_info(patch, pool, batch):
     """Print in terminal the patch size, the number of pooling and the batch size.
     """
+
     print("*"*20,"YOU CAN COPY AND PASTE THE FOLLOWING LINES INSIDE THE CONFIG FILE", "*"*20)
     print("BATCH_SIZE =", batch)
     print("PATCH_SIZE =", list(patch))
     aug_patch = get_aug_patch(patch)
-    print("AUG_PATCH_SIZE =",list(aug_patch))
+    print("AUG_PATCH_SIZE =",list(aug_patch))  
     print("NUM_POOLS =", list(pool))
 
 def auto_config(img_dir=None, median=None, max_dims=(128,128,128), max_batch=16, min_batch=2):
@@ -342,6 +362,8 @@ def auto_config(img_dir=None, median=None, max_dims=(128,128,128), max_batch=16,
         help="(default=\'configs/\') Configuration folder to save the auto-configuration.")
     parser.add_argument("--base_config", type=str, default=None,
         help="(default=None) Optional. Path to an existing configuration file which will be updated with the preprocessed values.")
+    parser.add_argument("--remote", default=False, dest='remote',
+        help="Use this arg when using remote autoconfing only")
     args = parser.parse_args()
 
     median = compute_median(path=args.img_dir, return_spacing=args.spacing)
@@ -353,8 +375,24 @@ def auto_config(img_dir=None, median=None, max_dims=(128,128,128), max_batch=16,
     patch, pool, batch = find_patch_pool_batch(dims=median, max_dims=(args.max_dim, args.max_dim, args.max_dim))
     aug_patch = np.array(patch)+2**(np.array(pool)+1)
 
-    display_info(patch, pool, batch)
-
+
+    if args.remote:
+        try: 
+            from biom3d.utils import save_python_config
+            config_path = save_python_config(
+                config_dir=args.config_dir,
+                base_config=args.base_config,
+                BATCH_SIZE=batch,
+                AUG_PATCH_SIZE=aug_patch,
+                PATCH_SIZE=patch,
+                NUM_POOLS=pool,
+            )
+            parameters_return(patch, pool, batch, config_path)   
+        except:
+            print("[Error] Import error. Biom3d must be installed if you want to save your configuration. Another solution is to config the function function in biom3d.utils here...")
+            raise ImportError
+    else :
+       display_info(patch, pool, batch)
     if args.spacing:print("MEDIAN_SPACING =",list(median_spacing))
     if args.median:print("MEDIAN =", list(median))
 
@@ -375,4 +413,4 @@ def auto_config(img_dir=None, median=None, max_dims=(128,128,128), max_batch=16,
             print("[Error] Import error. Biom3d must be installed if you want to save your configuration. Another solution is to config the function function in biom3d.utils here...")
             raise ImportError
 
-# ----------------------------------------------------------------------------
+# ----------------------------------------------------------------------------

src/biom3d/builder.py

@@ -695,7 +695,7 @@ def run_prediction_single(self, img_path=None, img=None, img_meta=None, return_l
                     **img_meta) # all img_meta should be equal as we use the same preprocessors
 
         else: # single model prediction
-            img, img_meta = read_config(self.config.PREPROCESSOR, register.preprocessors, img=img, img_meta=img_meta)
+            img, img_meta = read_config(self.config.PREPROCESSOR, register.preprocessors, img=img, img_meta=img_meta, is2d= self.config.IS_2D)
 
             print("Preprocessed shape:", img.shape)
 

src/biom3d/config_default.py

@@ -64,7 +64,7 @@ def __delattr__(self, name): del self[name]
 #---------------------------------------------------------------------------
 # Auto-config builder-parameters
 # PASTE AUTO-CONFIG RESULTS HERE
-
+IS_2D = False
 # batch size
 BATCH_SIZE = 2
 

src/biom3d/datasets/semseg_batchgen.py

@@ -76,6 +76,24 @@ def centered_crop(img, msk, center, crop_shape, margin=np.zeros(3)):
 
 def located_crop(img, msk, location, crop_shape, margin=np.zeros(3)):
     """Do a crop, forcing the location voxel to be located in the crop.
+
+    Parameters
+    ----------
+    img : ndarray
+        Image data.
+    msk : ndarray
+        Mask data.
+    location : array_like
+        Specific voxel location to include in the crop.
+    crop_shape : array_like
+        Shape of the crop.
+    margin : array_like, optional
+        Margin around the location.
+
+    Returns
+    -------
+
+        Cropped image and mask.
     """
     img_shape = np.array(img.shape)[1:]
     location = np.array(location)
@@ -96,6 +114,26 @@ def located_crop(img, msk, location, crop_shape, margin=np.zeros(3)):
 
 def foreground_crop(img, msk, final_size, fg_margin, fg=None, use_softmax=True):
     """Do a foreground crop.
+
+    Parameters
+    ----------
+    img : ndarray
+        Image data.
+    msk : ndarray
+        Mask data.
+    final_size : array_like
+        Final size of the cropped image and mask.
+    fg_margin : array_like
+        Margin around the foreground location.
+    fg : dict, optional
+        Foreground information.
+    use_softmax : bool, optional
+        If True, assumes softmax activation.
+
+    Returns
+    -------
+
+        Cropped image and mask.
     """
     if fg is not None:
         locations = fg[random.choice(list(fg.keys()))]
@@ -119,6 +157,20 @@ def foreground_crop(img, msk, final_size, fg_margin, fg=None, use_softmax=True):
 def random_crop(img, msk, crop_shape):
     """
     randomly crop a portion of size prop of the original image size.
+
+    Parameters
+    ----------
+    img : ndarray
+        Image data.
+    msk : ndarray
+        Mask data.
+    crop_shape : array_like
+        Shape of the crop.
+
+    Returns
+    -------
+
+        Cropped image and mask.
     """ 
     img_shape = np.array(img.shape)[1:]
     assert len(img_shape)==len(crop_shape),"[Error] Not the same dimensions! Image shape {}, Crop shape {}".format(img_shape, crop_shape)
@@ -135,6 +187,20 @@ def random_crop(img, msk, crop_shape):
 def centered_pad(img, final_size, msk=None):
     """
     centered pad an img and msk to fit the final_size
+
+    Parameters
+    ----------
+    img : ndarray
+        Image data.
+    final_size : array_like
+        Final size after padding.
+    msk : ndarray, optional
+        Mask data.
+
+    Returns
+    -------
+    tuple or ndarray
+        Padded image and mask, or only the image if mask is None.
     """
     final_size = np.array(final_size)
     img_shape = np.array(img.shape[1:])
@@ -156,6 +222,28 @@ def centered_pad(img, final_size, msk=None):
 def random_crop_pad(img, msk, final_size, fg_rate=0.33, fg_margin=np.zeros(3), fg=None, use_softmax=True):
     """
     random crop and pad if needed.
+
+    Parameters
+    ----------
+    img : ndarray
+        Image data.
+    msk : ndarray
+        Mask data.
+    final_size : array_like
+        Final size after cropping and padding.
+    fg_rate : float, optional
+        Probability of focusing the crop on the foreground.
+    fg_margin : array_like, optional
+        Margin around the foreground location.
+    fg : dict, optional
+        Foreground information.
+    use_softmax : bool, optional
+        If True, assumes softmax activation; otherwise sigmoid is used.
+
+    Returns
+    -------
+
+        Cropped and padded image and mask.
     """
     if type(img)==list: # then batch mode
         imgs, msks = [], []
@@ -179,6 +267,20 @@ def random_crop_pad(img, msk, final_size, fg_rate=0.33, fg_margin=np.zeros(3), f
     return img, msk
 
 class RandomCropAndPadTransform(AbstractTransform):
+    """
+    BatchGenerator transform for random cropping and padding.
+
+    Parameters
+    ----------
+    crop_size : array_like
+        Size of the crop.
+    fg_rate : float, optional
+        Probability of focusing the crop on the foreground.
+    data_key : str, optional
+        Key for the data in the data dictionary.
+    label_key : str, optional
+        Key for the label in the data dictionary.
+    """
     def __init__(self, crop_size, fg_rate=0.33, data_key="data", label_key="seg"):
         self.data_key = data_key
         self.label_key = label_key
@@ -601,6 +703,31 @@ def get_validation_transforms(patch_size: Union[np.ndarray, Tuple[int]],
 class BatchGenDataLoader(SlimDataLoaderBase):
     """
     Similar as torchio.SubjectsDataset but can be use with an unlimited amount of steps.
+
+    Parameters
+    ----------
+    img_dir : str
+        Directory containing the images.
+    msk_dir : str
+        Directory containing the masks.
+    batch_size : int
+        Size of the batches.
+    nbof_steps : int
+        Number of steps per epoch.
+    fg_dir : str, optional
+        Directory containing foreground information.
+    folds_csv : str, optional
+        CSV file containing fold information for dataset splitting.
+    fold : int, optional
+        Current fold number for training/validation splitting.
+    val_split : float, optional
+        Proportion of data to be used for validation.
+    train : bool, optional
+        If True, use the dataset for training; otherwise, use it for validation.
+    load_data : bool, optional
+        If True, loads the entire dataset into computer memory.
+    num_threads_in_mt : int, optional
+        Number of threads in multi-threaded augmentation.
     """
 
     def __init__(
@@ -810,6 +937,39 @@ def configure_rotation_dummyDA_mirroring_and_inital_patch_size(patch_size):
     return rotation_for_DA, do_dummy_2d_data_aug, initial_patch_size, mirror_axes
 
 class MTBatchGenDataLoader(MultiThreadedAugmenter):
+    """
+    Multi-threaded data loader for efficient data augmentation and loading.
+
+    Parameters
+    ----------
+    img_dir : str
+        Directory containing the images.
+    msk_dir : str
+        Directory containing the masks.
+    patch_size : array_like
+        The size of the patches to be extracted.
+    batch_size : int
+        Size of the batches.
+    nbof_steps : int
+        Number of steps per epoch.
+    fg_dir : str, optional
+        Directory containing foreground information.
+    folds_csv : str, optional
+        CSV file containing fold information for dataset splitting.
+    fold : int, optional
+        Current fold number for training/validation splitting.
+    val_split : float, optional
+        Proportion of data to be used for validation.
+    train : bool, optional
+        If True, use the dataset for training; otherwise, use it for validation.
+    load_data : bool, optional
+        If True, loads the entire dataset into computer memory.
+    fg_rate : float, optional
+        Foreground rate for cropping.
+    num_threads_in_mt : int, optional
+        Number of threads in multi-threaded augmentation.
+    """
+
     def __init__(
         self,
         img_dir,