TransPhy3D is a synthetic video dataset developed by the Beijing Academy of Artificial Intelligence in collaboration with the University of Southern California, Tsinghua University, and other institutions, focusing on transparent and reflective scenes. Related research papers include... Diffusion Knows Transparency: Repurposing Video Diffusion for Transparent Object Depth and Normal Estimation The dataset consists of 11,000 sequences rendered using Blender/Cycles, providing high-quality RGB frames along with physically based depth and normal labels.

Dataset composition:

• The dataset contains synthetic video sequences of transparent objects, and the scenes consist of richly categorized static assets and procedural assets with diverse shapes, which are paired with glass, plastic, and metal materials.
• The provided tags include RGB frame, depth, and normal tags, and the rendering is completed using physically based ray tracing and OptiX denoising.