High-resolution neural-network-based 3D imaging measures shape of moving objects using same projection patterns as for those stationary

https://www.eurekalert.org/news-releases/1107559

"using 3 standard sinusoidal projection patterns assumes object stationary... overcome using neural inverse rendering framework leveraging deep learning... models/ optimizes motion (represented as displacement field)/ shape, correcting for captured images' motion-induced discrepancies... extends structured light high-precision capabilities into dynamic environments merging deep learning/ optical principles... dynamic/ accurate 3D capture: manufacturing inspection, digital twin modeling, visual production performance"