Update modeling_fsg.py

modeling_fsg.py

@@ -351,37 +351,6 @@ class FsgModel(
     https://ieeexplore.ieee.org/abstract/document/9113265
 
     This class is designed to create a graph-based representation of forensic similarity between different patches of an image, allowing for the detection of manipulated regions.
-
-    Args:
-        stride_ratio (float): The ratio of the stride to the patch size, determining the overlap between patches. The lower the value, the higher the overlap.
-        fast_sim_mode (bool): If True, the algorithm uses a faster method to compute similarity scores, potentially at the cost of accuracy.
-        loc_threshold (float): The threshold for determining the location of interest in the similarity graph. Values above this threshold are considered significant.
-        is_high_sim (bool): If True, higher similarity scores indicate higher similarity. If False, lower scores indicate higher similarity.
-        need_input_255 (bool): If True, input images are expected to be scaled to [0, 255]. If False, images are expected to be in [0, 1].
-        **kwargs: Additional keyword arguments passed to the superclass initializer.
-
-    Example Usage:
-    ```python
-    import torch
-    import matplotlib.pyplot as plt
-    from torchvision.io import read_image, ImageReadMode
-    from model import FSG
-
-    ckpt_path = "path/to/ckpt.pth"
-    model = FSG.load_from_checkpoint(ckpt_path, map_location="cpu", stride_ratio=0.5, fast_sim_mode=False, loc_threshold=0.37, is_high_sim=False, need_input_255=False)
-    model.eval()
-
-    img_path = "path/to/image.jpg"
-    image = read_image(img_path, mode=ImageReadMode.RGB).float() / 255
-
-    device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
-    with torch.no_grad():
-        img_preds, loc_preds = model(image[None, ...].to(device))
-
-    plt.imshow(loc_preds.cpu()[0])
-    plt.colorbar()
-    plt.show()
-    ```
     """
     config_class = FsgConfig