A Hybrid Approach for Thermographic Imaging with Deep Learning

We propose a hybrid method for reconstructing thermographic images by combining the recently developed virtual wave concept with deep neural networks. The method can be used to detect defects inside materials in a non-destructive way. We propose two architectures along with a thorough evaluation that shows a substantial improvement compared to state-of-the-art reconstruction procedures. The virtual waves are invariant of the thermal diffusivity property of the material. Consequently, we can use extremely compact architectures that require relatively little training data, and have very fast loss convergence. As a supplement of the paper [1], we provide the MATLAB and Python implementations along with the data set comprising 40,000 simulated temperature measurement images in total, and their corresponding defect locations. Thus, the presented results are completely reproducible.