TransCG

From the catalogue. Listed by one researcher from one source page.

• id: transcg
• kind: grasping
• robot or device: RGB-D of transparent objects, robot-collected (class: sensor only)
• how collected: real capture with ground truth depth
• size as stated: 57,715 RGB-D images from 130 scenes
• hours: not stated | hours only claimed: none | episodes: not stated
• year: 2022
• organisation: SJTU MVIG
• licence: CC BY-NC-SA 4.0 (class: non commercial)
• access: open (class: open)
• link (dataset): https://graspnet.net/transcg
• note: Transparent object depth completion and grasping.

Confirm the licence at the link before relying on it.