Based on [CVPR 2020] 3D Photography using Context-aware Layered Depth Inpainting, we present image based 3D photograhpy.
There are four steps to generate 3D vidoe
Input: Single image S
(1) Geerate Depth image D from S
(2) Isolate foregroung mask F and background mask B from D
(3) Generate foreground layer F' and inpaiting background layer B'
(4) Generate images for 3D Video (TGA File)
(5) Used FFMpeg to generated MP4 video file
There are four directories in the repository.
(1) is Depth map generated, here we use MiDaS V3 to generate depth image from source image
(2) is the Foreground/Background isolate program: This will use histogram to decide best threshold for generate foreground and background.
(3) Generate Layer images: Use origin image, depth image,
