HIGHLIGHTS
SUMMARY
The deep learning methods encounter difficulty in material image processing. To overcome the above mentioned shortages, an interactive image segmentation is based on extracting center-environment features of each pixel. The center-environment features consist of domain knowledge based spatial information and several common texture features that can better represent rich and complex texture information in pixels and their adjacent pixels. For classifier, the authors choose some machine_learning algorithms to compare the intersection over Union (IoU), mean intersection over Union (mIoU), Accuracy, Dice Coefficient and mDice (mean Dice Coefficient) on many types of material . . .
If you want to have access to all the content you need to log in!
Thanks :)
If you don't have an account, you can create one here.