MultiDreamer3D: Multi-concept 3D Customization with Concept-Aware Diffusion Guidance
Authors: Wooseok Song, Seunggyu Chang, Jaejun Yoo
IJCAI 2025 | Venue PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Our experimental results show that Multi Dreamer3D not only ensures object presence and preserves the distinct identities of each concept but also successfully handles complex cases such as property change or interaction. To the best of our knowledge, we are the first to address the multi-concept customization in 3D. ... 4 Experiments |
| Researcher Affiliation | Collaboration | Wooseok Song1 , Seunggyu Chang2 and Jaejun Yoo1 1Ulsan National Institute of Science and Technology (UNIST) 2NAVER Cloud |
| Pseudocode | No | The paper describes the methodology in narrative text and mathematical formulations (e.g., in Section 3 "Method") but does not include any distinct pseudocode or algorithm blocks. |
| Open Source Code | No | The paper does not contain any explicit statement about releasing source code or a link to a code repository. |
| Open Datasets | Yes | We selectively choose real concept image data from the Custom Diffusion [Kumari et al., 2023] and Dream Booth [Ruiz et al., 2023] datasets, which contain 13 unique objects (three wearables and 10 unique objects). |
| Dataset Splits | No | The paper mentions using 13 unique objects from the Custom Diffusion and Dream Booth datasets and crafting 47 text prompts for evaluation, but it does not provide specific details on how these datasets were split into training, validation, or testing sets. |
| Hardware Specification | No | The paper does not provide specific details about the hardware used for running experiments, such as GPU or CPU models. |
| Software Dependencies | No | The paper mentions various models and frameworks like Shap-E, CLIP, 3DGS, and different diffusion models (e.g., Custom Diffusion, Dream Booth), but it does not specify version numbers for any of these software components or libraries. |
| Experiment Setup | No | The paper describes the overall method and evaluation, but it does not provide specific experimental setup details such as hyperparameter values (e.g., learning rate, batch size, number of epochs) or other system-level training configurations in the main text. |