This repo contains the subset of Objaverse used in LGM.
- kiuisobj_v1.txt: ~150K, filter bad prompts.
- kiuisobj_v1_merged_80K.csv: ~82K, v1 + filter missing textures. This is the subset used to train LGM.
- gobj_merged.json: ~83K, union of v1 and G-objaverse's 280K list.