Search

Challenge 3 was focused on the identification of defects present within graphene sheets. Provided with electron microscopy images of sheets of graphene, the data-set was partitioned into a sample of perfect patches (regions of an image which do not contain defects), defect patches (regions of an image which contain a defect) and a data-set of images which are not edited or partitioned into smaller sections of analysis. The full image is 256 x 256 patches (Figure 1a). The blue (high electron density) corresponds to atoms and the green corresponds to background. In the full image patches, there is a perfect 48 x 48 patches and a defect 48 x 48 patches. By selecting and training an appropriate machine learning model, the goal was the identification of defect regions contained within a whole electron microscopy image of a graphene sheet.