Solid Wood Board Dataset
We established a data set about solid wood board, which was used for image inpainting in order to reconstruct texture. These images are regular small-size images segmented from the complete large image, so as to be used as the input of the model.
Steps to reproduce
We built an automated transport platform for taking photos of the surface texture of solid wood panels, equipped with an OscarF810CIRF industrial camera. The captured photos are cropped to 200×200 pixels and constitute the data set for model training and testing. In order to better fit our model, we randomly selected 80% of the original data set as the training set. Then the original training set was expanded to six times the original through four expansion methods. The first method was used to mirror the upper and lower parts of all images in the training set with the horizontal axis of the image as the symmetry axis; the second method was used to mirror the left and right parts of all images in the training set with the vertical axis of the image as the symmetry axis; The third method randomly extracted one-half of the original training set and applied a random brightness transformation to it; The fourth method randomly extracted one half of the original training set and performed a random contrast transformation on it. The remaining 20% of the original data set was used as the test set of the model.