MEPhI National Research Nuclear University has presented a new neural network, 3D-CGH-Net, for creating holographic images. The solution will replace traditional resource-intensive methods for calculating diffraction elements for the formation of three-dimensional scenes.
The development accelerates the process of creating optical elements, providing high quality holographic scenes. Previously, similar results could only be achieved with more complex and slower methods.
We have developed a method that uses a neural network of original architecture and a branched structure to account for a large set of sections of a three-dimensional scene in the calculated hologram. The network is trained on samples ranging from tens of thousands to hundreds of thousands of examples.
To create computer-generated holograms, it is necessary to calculate the shape of the diffraction element based on the required distribution of amplitude and phase of light. Existing "classical" methods are iterative and computationally intensive. In modern conditions, this is unacceptable.
With 3D-CGH-Net, the process of calculating holograms becomes faster. This opens up new opportunities for the use of holography in optics, medical imaging, and virtual reality.
Earlier, www1.ru reported that a project to analyze tomography for cancer diagnosis has started at NSU.
Read materials on the topic:
Russia wants to train AI to create cancer vaccines
AI-based Russian Sign Language translator created in Russia
AI helped reduce mortality from vascular diseases in a Khabarovsk hospital with over 95% accuracy