We present a new part-based 3D generation framework, CoPart, which represents a 3D object with multiple contextual part latents and simultaneously generates coherent 3D parts. And we are pleased to release the first large-scale 3D object part dataset PartVerse that has been manually annotated.

We follow the pipeline of “raw data - mesh segment algorithm - human post correction - generate text caption” to produce part-level data.

You can download the PartVerse dataset from Google Drive, Huggingface. The data directory after decompressing the dataset should be as follows:

dataset/
├── textured_part_glbs/  
├── normalized_glbs/    
├── anno_infos/    
└── text_captions.json       

• textureld_part_glbs contains textured 3D meshes for each decomposed part of the objects. Each file is stored in the GLB format.
• normalized_glbs provides the complete, normalized 3D mesh of each object in GLB format. These are aligned with the part-level meshes and can be used for holistic shape analysis or comparison.
• anno_infos provides files that can be used for generating auxiliary information of parts.
• text_captions.json stores descriptive text captions for each part, automatically generated using a Vision-Language Model (VLM).

Due to the large number of parts in some objects, we can discard some unimportant parts (such as a screw, etc.). We provide partverse/get_infos.py to process the data. By running it, you can obtain (1) some statistical information of the parts, (2) the priority of discarding them, (3) view of max overlap between full object and parts render. Please install nvdiffrast and kaolin when use.

python partverse/get_infos.py --data_root ${DATA_PATH} --global_info_save_path ${SAVE_PATH} --max_visible_info_save_path ${SAVE_PATH}

We provide rendering script following TRELLIS. You can use partverse/render_parts.py to render textured_part_glbs (part objects) and partverse/render_dir.py to render normalized_glbs (whole objects), e.g.,

python partverse/render_parts.py --textured_part_glbs_root ${PART_GLB_PATH} --out_dir ${OUT_PATH} --num_views 8 --elevation 30

In addition, we also provide text caption code to facilitate users in customizing text prompts for their own models. For the VLM, we use Qwen2.5-VL-32B now. You can replace to any VLM.

python partverse/get_text_caption.py --raw_img_root ${FULL_OBJECT_IMG_PATH} --part_img_root ${PART_IMG_PATH} --info_file ${MAX_VIS_INFO_PATH} --output_file ${OUT_PATH} --vlm_ckpt_dir ${VLM_HF_DOWN_PATH}

• [2025/07/05] Our paper is accepted by ICCV 2025. Code is coming soon, stay tuned! 🔥

@article{dong2025copart,
  title={From One to More: Contextual Part Latents for 3D Generation},
  author={Shaocong Dong, Lihe Ding, Xiao Chen, Yaokun Li, Yuxin WANG, Yucheng Wang, Qi WANG, Jaehyeok Kim, Chenjian Gao, Zhanpeng Huang, Zibin Wang, Tianfan Xue, Dan Xu},
  booktitle={ICCV},
  year={2025}
}