Sculpture Detection Method using the Convolution Neural Network
Dajeong Hong, Jongweon Kim · 2017
In this paper, we propose a method for detecting a sculpture in an image using CNN. The sculpture is made of various materials unlike other works of art. It is not easy to perceive it because it may look slightly different in each the image according to various conditions around it. In our method, an image containing a part of a sculpture or an image of a sculpture be taken at various angles is collected. Then, we create a dataset by setting the location of the sculpture in each image to ROI (Region of Interest). We trained datasets using SSD networks with high accuracy and speed. After, with the learned model, detected and recognized the sculptures in the test images. As a result, obtained about 82% accuracy.