Mastering Computer Vision with TensorFlow 2.x

Overview of hardware design for 3D reconstruction

3D reconstruction involves cameras, sensors, lighting, and depth estimation. The sensors used in 3D reconstruction can be divided into three categories:

• Multiview setups: A calibrated dense stereo camera array with controlled illumination. From each stereo pair, the face geometry is reconstructed using triangulation and then aggregated while enforcing geometric consistency.
• RGB camera: Multiple RGB cameras are combined to calculate depth based on the time-of-flight approach.
• RGBD camera: RGBD camera captures both color and depth—some examples are Microsoft Kinect, Primesense Carmine, and Intel Realsense.