Deep learning to detect built cultural heritage from satellite imagery. - Spatial distribution and size of vernacular houses in Sumba, Indonesia -

In Sumba Island – Indonesia, the implantation of vernacular houses, inside and outside traditional villages, is considered to be an efficient proxy for on-going complex cultural transformations resulting from globalization. This study presents easily reproducible workflow allowing buildings automatically detected satellite imagery, demonstrating how modern computer vision methods based on deep learning can help in this task, which would far too time-consuming when undertaken by hand. Eight architectures convolutional neural networks were compared terms ability identify locate precisely houses images. By combining a Faster R-CNN ResNet 101 architecture with artificial data augmentation, model was taught properly using 1033 instances ([email protected]:.95 = 71.9). Once 14 952 had been detected, Histogram Orientated Gradients (HOG) computed processed several machine algorithms assess their surface area, as parameter conveys pertinent information about economic political position householder. The best classifier found support vector regressor (SVR, R2 0.88), although other classifiers tested also provided good results (R2 > 0.76). Spatial analysis used draw conclusions anthropological / identity point view. More generally, these techniques not only offer simple increase recording capabilities tangible heritage, they open up new research perspectives, at greater scales.