Crack detection has vital importance for structural health monitoring and inspection. We would like to train a network to detect Cracks, we will denote the images that contain cracks as positive and images with no cracks as negative.
This model applies transfer learning on Resnet-18 pre-trained model in order to classify the input images of concrete as positive or negative.
To download data run data.sh in data folder or manually download from Positive_tensor.zip & Negative_tensor.zip
Plot of loss during training on given data:
Capstone project for IBM AI Specialization