fix(image-utils): fix all functions

Author: Vinschers

VERSION

@@ -1 +1 @@
-0.13.2
+0.14

src/vicentin/deep_learning/train/GANTrainer.py

@@ -23,7 +23,7 @@ def __init__(
         noise_dim: int = 100,
         n_imgs_val: int = 64,
         sample_path: str = "gan_samples",
-        export_mode: Literal["save", "show", None] = "show",
+        export_mode: Literal["save", "show", None] = None,
         fixed_z: Optional[torch.Tensor] = None,
         hyperparams: Optional[Dict[str, Any]] = None,
         device: Optional[torch.device] = None,

src/vicentin/deep_learning/train/VAETrainer.py

@@ -21,7 +21,7 @@ def __init__(
         float16: bool = False,
         sample_path: str = "vae_samples",
         n_imgs_val: int = 8,
-        export_mode: Literal["save", "show", None] = "show",
+        export_mode: Literal["save", "show", None] = None,
     ):
         super().__init__(
             models={"vae": model},

src/vicentin/image/__init__.py

@@ -1,3 +1,4 @@
+from . import utils
 from . import differentiation
 from . import regularization
-from . import utils
+from . import video

src/vicentin/image/img2graph.py

@@ -1,8 +1,11 @@
+from typing import Any, Callable, Optional
 from vicentin import Graph
 from vicentin.image.utils import get_neighbors
 
 
-def _compute_edges(H, W, L, img, weight_func, neighborhood):
+def _compute_edges(
+    H: int, W: int, L: int, img: Any, weight_func: Callable, neighborhood: int
+):
     """
     Computes the weighted edges of a graph representation of an image.
 
@@ -14,7 +17,7 @@ def _compute_edges(H, W, L, img, weight_func, neighborhood):
         H (int): The height (number of rows) of the image.
         W (int): The width (number of columns) of the image.
         L (int): The depth (number of channels) of the image.
-        img (numpy.ndarray or jnp.ndarray): A 1D, 2D, or 3D NumPy array representing the image.
+        img (numpy.ndarray or torch.Tensor): A 1D, 2D, or 3D NumPy array representing the image.
         weight_func (callable): A function `f(u, v) -> float` that computes edge
                                 weights based on pixel values.
         neighborhood (int): The type of neighborhood connectivity:
@@ -51,7 +54,9 @@ def _compute_edges(H, W, L, img, weight_func, neighborhood):
     return edges
 
 
-def img2graph(img, weight_func=None, neighborhood=4):
+def img2graph(
+    img: Any, weight_func: Optional[Callable] = None, neighborhood: int = 4
+):
     """
     Converts a 1D, 2D, or 3D image into a graph representation.
 
@@ -60,7 +65,7 @@ def img2graph(img, weight_func=None, neighborhood=4):
     are determined by a user-defined function (defaulting to squared intensity difference).
 
     Args:
-        img (numpy.ndarray or jnp.ndarray): An array of shape (N,), (N, M), or (N, M, L)
+        img (numpy.ndarray or torch.Tensor): An array of shape (N,), (N, M), or (N, M, L)
                              representing a 1D, 2D, or 3D image.
         weight_func (callable, optional): A function `f(u, v) -> float` that computes edge
                                           weights based on pixel values.

src/vicentin/image/utils/__init__.py

@@ -0,0 +1,8 @@
+from .utils import (
+    pad_image,
+    correlate,
+    convolve,
+    gaussian_filter,
+    img2patches,
+    get_neighbors,
+)