Researchers have long sought to create computer-generated holograms, but the process
“Previously, scientists thought that with the existingconsumer-grade hardware cannot perform real-time 3D holography computations, ”said Liang Shi, lead author of the study and a graduate student in the Department of Electrical and Computer Science (EECS) at the Massachusetts Institute of Technology (MIT).
Shi believes the new approach, which the team calls "tensor holography", will help achieve this goal faster. This will help create holography in 3D and VR.
They used deep learning to speed upcomputer holography, allowing them to be generated in real time. The team designed a convolutional neural network, a processing technology that uses a chain of trainable tensors to roughly mimic how a person processes visual information. Training a neural network usually requires a high quality dataset that did not previously exist for 3D holograms.
The team has created a custom database,consisting of 4 thousand pairs of images generated by a computer. To create holograms in the new database, the researchers used scenes with complex and variable shapes and colors, with uniform pixel depths from background to foreground, and a new set of physics-based calculations to handle occlusion.
Learning on each pair of images, tensorthe network tweaked the parameters of its own calculations, consistently improving its ability to create holograms. The fully optimized network ran an order of magnitude faster than physical calculations.
Read more
Physicists have created an analogue of a black hole and confirmed Hawking's theory. Where it leads?
Sea slugs cut off their head to regenerate a new body
Abortion and science: what will happen to the children who will give birth