Fixed: Dimension error when looping through multiple images with image-specific point coordinates

[jasperkleinsmann]

I used model.generate_masks_by_points_patch() to loop through multiple images with their own set of xy coordinates. The first iteration would run fine, but an error would occur when running the next iteration. I found that this happened because neither model.generate_masks_by_points_patch() or model.set_image() resets the argument model.images_batch to None. Hence in my case, the model took some image parameters from the previous iteration but other image parameters from the current iteration causing a dimension mismatch. I am sure there is a better way to fix this but in my case it was solved by manually setting model.images_batch = None.

Just sharing this in case someone else runs into a similar issue.

for i, img in enumerate(images_paths):
 
    # Reset model (images_batch is not automatically reset to None causing errors when looping)
    sam3.images_batch = None
    
    # Import image and xy coordinates
    image_path = image_folder + img
    pnt_coords_image = read_clip_vector(pnt_coords_path)

    # Specify image 
    sam3.set_image(image_path) 
    
    # Segment image based on point coordinates
    sam3.generate_masks_by_points_patch(
        point_coords_batch = pnt_coords_image,
        point_crs="EPSG:25832",
        output=f'{output_folder}/mask_{i}.tif',
        dtype="uint8"
    )

[giswqs]

Thank you for reporting. The issue has been fixed in #495.