At the moment, deep convolutional networks are state-of-the-art algorithms in many problems of computer vision. However, most of the algorithms require huge training datasets, and the quality of the model depends entirely on the quality and quantity of the data. In a number of tasks, data collection is difficult and sometimes impossible. In the talk, we will consider an example of training deep convolutional networks for object keypoints localization on a fully synthetic data set.
Download presentation