Deep structure learning using feature extraction in trained projection space

Over the last decade of machine learning, convolutional neural networks have been most striking successes for feature extraction rich sensory and high-dimensional data. While learning data representations via convolutions is already well studied efficiently implemented in various deep libraries, one often faces limited memory capacity insufficient number training data, especially large-scale tasks. To overcome these limitations, we introduce a network architecture using self-adjusting dependent version Radon-transform (linear projection), also known as X-ray projection, to enable lower-dimensional space. The resulting framework, named PiNet, can be trained end-to-end shows promising performance on volumetric segmentation We test proposed model public datasets show that our approach achieves comparable results only fractional amount parameters. Investigation usage processing time confirms PiNet’s superior efficiency compared other models.