Can you imagine creating a detailed, realistic 3D face model with a single click of a portrait?

Through a large number of sample data collected from its own 3D face database, the research team of VisionChina built neural network training, and finally made a breakthrough in the key technology of recovering 3D model from 2D face image by deep learning.

Zhou Xiang, president and general manager of Xi 'an Zhixiang Optoelectronics, told us: "This technology supports obtaining a high-resolution depth image from a single two-dimensional portrait image, and then recovering 3D point cloud data to output a realistic 3D face model." VisionChina Shanghai this year will present its innovative machine vision products.

Xi 'an Zhixiang Photoelectric Chairman and General Manager Zhou Xiang

Using this technology to restore the 3D face model has a very real color and accurate size, has a good visual experience. It can help users break through the limitations of two-dimensional images, reduce the cost of modeling, improve the efficiency of modeling, and has a very broad application prospect in 3D printing, medical plastic surgery, game entertainment, video communication and other fields.

3D portrait modeling solution

The 3D face restoration algorithm can automatically generate a 3D portrait model that highly matches the person in the picture through a single portrait photo taken by an ordinary camera. In addition, the AI intelligent completion algorithm independently developed by Zhixiang Optoelectronic can also intelligently match and integrate the collected 180-degree (left ear to right ear) 3D face data with its own head information database data to achieve complete head reconstruction. Based on the head 3D model and the photoelectric image own portrait model library, hair library can also generate customized 3D model according to customer requirements.

Zhixiang photoelectric machine vision products

Surface series industrial 3D cameras use ultra-fast laser-encoded full-field projection mode, and can output 3D point clouds of 2~15 Hz, with repeatability accuracy up to 0.05mm. The independent AI chip is used for embedded calculation, and the coordinate point data can be output directly without connecting to the high-configuration computer. Using infrared laser light source, strong anti-environmental interference, lightweight and easy to integrate, can be competent for workpiece alignment, path tracking, palletizing, loading and unloading and other tasks.

Surface Industrial 3D Camera

The 3D camera Ailook adopts independent MEMS binocular technology, with an accuracy up to 0.1mm. It supports real-time dynamic scanning and output RGB-D images of 2~15FPS. Through the national metrological certification; Using safe non-inductive infrared light source, embedded computing, multi-platform compatibility, with a high degree of integration.

Ailook3D camera

Independent research and development of a new generation of portable 3 d scanner, the non-inductive infrared light source control technology and artificial intelligence algorithms, high-performance embedded computing ability, the fuselage clever and compact, portable and easy to use, can complete a variety of complex efficient scanning task, also can be used to scan modeling of face, body and meet the needs of product development and prototyping, high cost performance, Make the scanning process more convenient and time saving, and make the 3D data more accurate and reliable.

HandyLook, a handheld 3D scanner