yaredtaddese/leaf-image-segmentation

Leaf Image Segmentation

leaf-image-segmentation's Issues

Crash when no leaf is present

Yuca plant root.

Traceback (most recent call last):
  File "leaf-image-segmentation/segment.py", line 163, in <module>
    segment_leaf(os.path.join(base_folder, file), filling_mode, smooth, args.marker_intensity)
  File "leaf-image-segmentation/segment.py", line 66, in segment_leaf
    select_largest_obj(bin_image, fill_mode=filling_mode,
  File "leaf-image-segmentation/background_marker.py", line 213, in select_largest_obj
    largest_obj_lab = np.argmax(lab_stats[1:, 4]) + 1
  File "<__array_function__ internals>", line 5, in argmax
  File "leaf-image-segmentation/venv/lib/python3.8/site-packages/numpy/core/fromnumeric.py", line 1188, in argmax
    return _wrapfunc(a, 'argmax', axis=axis, out=out)
  File "leaf-image-segmentation/venv/lib/python3.8/site-packages/numpy/core/fromnumeric.py", line 58, in _wrapfunc
    return bound(*args, **kwds)
ValueError: attempt to get argmax of an empty sequence

Crash on flood fill

Traceback (most recent call last):
  File "leaf-image-segmentation/segment.py", line 162, in <module>
    segment_leaf(os.path.join(base_folder, file), filling_mode, smooth, args.marker_intensity)
  File "leaf-image-segmentation/segment.py", line 66, in segment_leaf
    select_largest_obj(bin_image, fill_mode=filling_mode,
  File "leaf-image-segmentation/background_marker.py", line 229, in select_largest_obj
    cv2.floodFill(img_floodfill, mask_, seedPoint=bkg_seed,
cv2.error: OpenCV(4.5.1) /tmp/pip-req-build-ms668fyv/opencv/modules/imgproc/src/floodfill.cpp:509: error: (-211:One of the arguments' values is out of range) Seed point is outside of image in function 'floodFill'

Image:

How to give input in file segment.py

I have never used argparse library, so I am having difficulty modifying the code. Instead of giving the image source while running the python file that is python segment.py testimage, I want give the path of image inside the segment.py file . How to do it?

Lesion on edge of leaves is ignored when segmented.

Lesion on edge of leaves is ignored when segmented, practically denying that the lesion is part of the leaf and considering it as background rather than leaf part. This results in a bad segmentation for applications that rely on the existence of such lesion parts of leaves in the input, for example, for applications that detect plant diseases and the likes.