Tool crashes on loading Pre-Annotated Images about js-segment-annotator HOT 15 CLOSED

@kyamagu Thank you very much. It worked. :)

from js-segment-annotator.

@anandcu3 Do you change the set of labels? The crash happens because the annotation data seem to contain a label index outside of the given set.

https://github.com/kyamagu/js-segment-annotator/blob/master/js/app/edit.js#L476

from js-segment-annotator.

@kyamagu I did not change the labels but the annotation image was generated by a inaccurate ML algorithm . But I made sure the value in the annotations are between 0 and n (num of classes). 1->Sky, 2->Building etc..

from js-segment-annotator.

@anandcu3 You have to have 0 index (usually background) in the label set.

from js-segment-annotator.

@kyamagu Isn't that Sky in my case ?

from js-segment-annotator.

@anandcu3 You are mapping Sky to 1, not 0.

from js-segment-annotator.

If you really intend to use Sky for 0, then I assume there is invalid integer index in the incomplete data.

from js-segment-annotator.

@kyamagu I'm sorry I made a mistake in my earlier message. I have mapped sky-->0, building->1 and so on.. So background is sky which has value 0 in the image.

from js-segment-annotator.

@anandcu3 Ok, then first check unique set of index values in the annotation PNG.

from js-segment-annotator.

The other question is are there any specifications to the annotations image format that has to be taken care of ..

from js-segment-annotator.

@anandcu3 The format is specified in README.

from js-segment-annotator.

@kyamagu

a=cv2.imread("data/annotations/00001.png")
print(numpy.unique(a))

returned

array([ 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10], dtype=uint8)

from js-segment-annotator.

@anandcu3 The PNG is RGB-encoded UINT32 and cannot be read just by cv2.imread. Make sure all the index values are properly formatted. Other than that, you can always open up browser javascript console to inspect values.

from js-segment-annotator.

@kyamagu I've an array with the same shape of an image. Each pixel representing the class label. How do I encode this? I've been trying to use the encode portion of the code in README but I get this error.

Traceback (most recent call last): File "D:\anand\anaconda3\lib\site-packages\PIL\Image.py", line 2167, in fromarray mode, rawmode = _fromarray_typemap[typekey] KeyError: ((1, 1, 3, 3), '|u1')

During handling of the above exception, another exception occurred: Traceback (most recent call last): File "<stdin>", line 1, in <module> File "D:\anand\anaconda3\lib\site-packages\PIL\Image.py", line 2170, in fromarray raise TypeError("Cannot handle this data type") TypeError: Cannot handle this data type

from js-segment-annotator.

@anandcu3 Convert the array to uint32 first, then try encoding with the code example.

from js-segment-annotator.