## 📝 Challenge Rules

Participants are required to submit all **private `.obj` mesh files** corresponding to the selected private food items. These submissions will be evaluated based on **shape and size accuracy** of the reconstruction.

Please carefully review the following rules:

### ✅ Submission Guidelines

- Submissions **must follow the provided instructions**. Submissions will be disqualified if they contain inappropriate content or do not meet the official rules.
- Each registered participant or team may **submit once per day**.
- The contest is **open to all**. However, individuals involved in organizing or supporting this competition are **not eligible to participate**.
- Participants may use **publicly available data and models**, but **must provide proper attribution** by including references in their submitted code.
- During the **final evaluation phase**, the organizers will thoroughly inspect the submitted code and 3D models of top-performing teams.  
  Disqualification will occur in the following cases:
  - Use of plagiarized or similar code to other teams.
  - Submissions lacking substantial original work.
  - Submissions with 3D models that do **not represent the correct shape of the object class**.  
    For example:  
    ❌ Submitting a **sphere** for the class `strawberry`  
    ✅ Submitting a **hemisphere with a cone** for the class `strawberry`
- Mesh does not have to be watertight.
- Contestants are fully responsible for ensuring their submissions, including files, data, and code, are **legal and comply with all regulations**.  
  The organizers **assume no legal liability** for issues that arise from participant submissions.
- If participants publish papers related to their results in this competition, **they must acknowledge the competition**.

### 🏆 Winner and Prize

- The **top-3 teams** will be selected as winners.
- Winners will be **invited for an oral presentation** at the MetaFood workshop and will be asked to write a **technical report** about the challenge.
- Visit the [MetaFood 2025 Workshop Website](https://sites.google.com/view/cvpr-metafood-2025) for prize details.

---

## 📦 Use of External Resources

Participants are allowed to use **external resources**, provided **sources are clearly cited** in the submission. Here are some external resources you can start with.

### 3D Generation Models
- TRELLIS [[1]](#references)
- Hunyuan3D 2.0 [[2]](#references)

### 3D Object Datasets
- OmniObject3D [[3]](#references)
- NutritionVerse-3D [[4]](#references)
- MetaFood3D [[5]](#references)
- Objaverse [[6]](#references)

### Depth Estimation Models
- DepthPro [[7]](#references)
- Depth Anything V2 [[8]](#references)

---

## 📚 References

<a name="references"></a>

[1] Xiang, Jianfeng, et al. "Structured 3D Latents for Scalable and Versatile 3D Generation." *arXiv preprint arXiv:2412.01506* (2024).  
[2] Zhao, Zibo, et al. "Hunyuan3D 2.0: Scaling Diffusion Models for High Resolution Textured 3D Assets Generation." *arXiv preprint arXiv:2501.12202* (2025).  
[3] Wu, Tong, et al. "OmniObject3D: Large-Vocabulary 3D Object Dataset for Realistic Perception, Reconstruction and Generation." *CVPR 2023*.  
[4] Tai, Chi-en Amy, et al. "NutritionVerse-3D: A 3D Food Model Dataset for Nutritional Intake Estimation." *arXiv preprint arXiv:2304.05619*, 2023.  
[5] Chen, Yuhao, et al. "MetaFood3D: Large 3D Food Object Dataset with Nutrition Values." *arXiv preprint arXiv:2409.01966* (2024).  
[6] Deitke, Matt, et al. "Objaverse-XL: A Universe of 10M+ 3D Objects." *NeurIPS 2023*, pp. 35799–35813.  
[7] Bochkovskii, Aleksei, et al. "DepthPro: Sharp Monocular Metric Depth in Less Than a Second." *arXiv preprint arXiv:2410.02073* (2024).  
[8] Yang, Lihe, et al. "Depth Anything V2." *NeurIPS 2024*, pp. 21875–21911.

---