Abstract: “Deep neural networks (DNNs) are used to reconstruct transmission speckle intensity patterns from therespective reflection speckle intensity patterns generated by illuminated parafilm layers. The dependence ofthe reconstruction accuracy on the thickness of the sample is examined for different illumination patterns ofvarious feature sizes. High reconstruction accuracy is obtained even for large parafilm thicknesses, for whichthe memory effect of the sample is vanishingly small. The generalization capability of the DNN is also studiedfor unseen scatterers of the same type.”