Segmentation of mine overburden dump particles from images using Mask R CNN

As a result, the shadows of dump particles affected a few neighboring particles. Dump image dataset was fed to the model without applying any morphological operations so that the model can identify dump particles from in situ condition. In this study, Mask R CNN has been implemented using ResNet50 with FPN. The same multi task loss function as defined in original Mask R CNN paper23 has been implemented. $$\begin{aligned} L=L{cls}+L{bbox}+L_{mask} \end{aligned}$$ (1)For implementing the Mask R CNN, the dump image dataset was split into three parts.