BinRushed dataset have total 195 original images . There are annoted image by 6 different experts . If I have done segmentation of original using a method then how can I check for accuracy of my implemented method?
In that remove the other annotated images from dataset and oversample your data while training your model. but if you cannot train again than just give your data as test data and create confusion matrix against data annotated by others.
Hope this help!
Have a look at Article Machine printed character segmentation —; An overview
and https://stackoverflow.com/questions/13974167/how-to-test-accuracy-of-segmentation-algorithm
You can check segmented image accuracy by calculating the Dice score by calculating the intersection between the segmented image and the ground truth segmentation and dividing on the whole area.
But I don't have ground Truth images for checking
Mahmoud Khaled Abd-Ellah
Dear Shalaka Jagtap
You can create it with an expert. Also, you can search for other datasets.
- Badges
- Science topic
- Similar topics
- Computer Science
- Images