An efficient model for semantic segmentation of micro-structure images to detect various surface defects. This model is implemented using TensorFlow and Keras. The network architecture is inspired by the dcgan model and uses UNET as a generator.
Raw Dataset: https://dasci.es/transferencia/open-data/metal-dam/
the altered dataset and the best model Checkpoint so far: https://drive.google.com/drive/folders/1ZOAIyTEc4OQCIr_N5-EQ_PkGzS4CaeV0