Group 5: Challenge: Nanopore “Defect detection in graphene sheets”

Graphene is a nanomaterial with excellent super conducting properties, is considered to be the strongest material available and is impermeable to gasses (even Helium-the smallest gas atom). Its properties find applications in fabrication of electronic and optoelectronic devices, gas sensors such as chemiresistors for detection of ammonium [1] and nitrogen dioxide [2], biosensors, composite materials and energy storage devices. Graphene has a regular structure- similar to a honey comb, which consists of fused six membered carbon rings into 2D array. This atomic structure is the foundation of the exceptional properties of the material. The occurrence of defects in the lattice which occur during the manufacturing process can cause significant deterioration in the property of the final material produced [3]. These defects can be detected with a number of spectroscopy-microscopic techniques, one of which is Surface Scanning Electron Microscopy (SEM). In the challenge Nanopore “Defect Detection in Graphene Sheets” a complete dataset of 180 SEM images 256x256 pixels (full-stack), from which were derived 2279 images 48x48 pixels of perfect patches (pp) and 32 images 48x48 pixels of defect containing patches (dp) were provided in numpy arrays, see Figure 1. All pixel values had been normalised to values between 0.00 and 1.00. Figure 1 Nanopore “Defect Detection in Graphene Sheets”. The task was to design a classifier for SEM graphene images using the pp dataset. The two classes are as follows: the first is images without defects and the second class, images with defects, where the type of defect or its localisation are not considered. Finally, the classifier had to be evaluated using an appropriate metric on the dp dataset