美图
三维视觉算法实习生（图像 / 视频美型）
三维视觉算法实习生（图像 / 视频美型）
发布于 大约 8 小时前实习/见习
澳大利亚
无经验要求
实习生
远程工作
博士
实习与临时职位
PyTorch
AIGC
Diffusion Models
Smpl
Human Pose Estimation
AI 估算 · 15k–25k
PhD intern in Australia, remote, competitive AI field, estimated monthly stipend in RMB.
职位详情
关于这个职位
This internship focuses on 3D vision and human body beautification algorithms, involving tasks like 2D/3D human tracking, 3D reconstruction, and controllable editing of images/videos. You will work on cutting-edge AI for portrait enhancement and temporal consistency, with opportunities to publish research papers and deploy real-world products.
最低要求
PhD candidate in Computer Science, AI, Intelligent Art, or related fields
Proficient in Python with common libraries (numpy, transformer, diffuser, open3d), strong deep learning experience (PyTorch)
Good engineering implementation skills and experimental habits, able to complete module-level tasks independently
Proficient in using AI programming tools, good code discipline and organization habits
Good ability in paper reading, understanding, analysis, concept summarization, and code reproduction
Logical thinking, good at problem decomposition, willing to deeply understand the essence of problems
Long-term interest in AI + imaging / 3D / aesthetics
Proactive, willing to explore new methods and embrace challenges
Identify with a product culture oriented towards user experience and technology value
工作职责
Participate in testing and evaluation of existing CNN-based beautification methods
Participate in fine-tuning applied algorithms, training, and experimental result analysis
Improve algorithm networks, including design of beautification detail methods (local deformation/body shape correction)
Provide algorithm demos, collaborate with product and other R&D teams to ensure algorithm effectiveness and drive project deployment
Later stage may participate in AIGC big data generation projects such as diffusion models
优先资格
Experience with SMPL / human pose estimation / 3D Reconstruction / Tracking
Familiarity with any of: video understanding / diffusion models / autoregressive / flow matching models
Experience in research projects, competitions, or open-source projects
Strong mathematical or geometric intuition, good aesthetic sense, passionate about excellent design
AI 洞察
优缺点分析
优点
- Work on cutting-edge 3D vision and generative AI with real product impact.
- Access to large-scale user data and compute resources.
- Opportunity to publish research papers alongside product work.
- Remote-friendly and flexible working environment.
- High expectations for independent research and engineering output.
- Need to balance research novelty with product constraints.
- Competitive field requiring continuous learning of new methods.
- This role is ideal for PhD students passionate about 3D vision and human-centric AI, who enjoy both research and engineering and seek to make a tangible impact on consumer products.
缺点 / 挑战
暂无明显挑战项
角色解读
- Deepen expertise in 3D vision and human modeling, potentially leading to a full-time research role.
- Opportunity to publish high-quality papers at top conferences.
- Transition to product-focused roles in AI imaging or continue as a researcher in academia/industry.
- Test and evaluate existing CNN-based beautification methods for human images/videos.
- Fine-tune and train algorithms for 2D/3D human tracking and 3D reconstruction.
- Design improved network architectures for local deformation and body shape correction.
- Provide algorithm demos and collaborate with product teams for deployment.
- Strong Python programming and deep learning (PyTorch) skills.
- Familiarity with libraries like numpy, transformer, diffuser, open3d.
- Ability to read and reproduce research papers independently.
- Proactive problem-solving and logical thinking.
申请策略
- Tailor your application to show alignment with Meitu's mission of 'beauty meets technology'.
- Prepare a brief portfolio of relevant work (papers, code, demos) to share.
- Showcase prior projects or papers in 3D reconstruction, human pose estimation, or diffusion models.
- Demonstrate strong coding skills with relevant GitHub repos or open-source contributions.
- Emphasize experience with PyTorch and any practical deployment of AI models.
- Highlight any awards, competitions, or research internships in related fields.
- Strengthen knowledge of SMPL model and human body parametric models.
- Get familiar with state-of-the-art diffusion models for image/video generation.
面试指南
- Use STAR (Situation, Task, Action, Result) to structure answers about past projects.
- For technical questions, explain the problem, compare alternative approaches, and justify your choice.
- Show enthusiasm for the intersection of research and product by discussing how you balance innovation with practicality.
- Describe your experience with 3D reconstruction or human pose estimation. What methods have you used?
- How would you approach fine-tuning a CNN for body shape correction while maintaining temporal consistency?
- Explain the concept of SMPL and how it can be used for human beautification.
- What are the challenges in applying diffusion models to video generation?
- How do you ensure code quality and reproducibility in your research projects?
职位点评
68
综合评分
Cutting-edge 3D vision intern with strong research focus, remote flexibility, and high growth potential, but unclear compensation and no explicit WLB guarantees.
更适合这类人
This role is best for PhD candidates who prioritize skill growth and research impact over immediate compensation or WLB.
表现最好
成长发展
相对薄弱
薪资福利
薪资福利50
成长发展90
工作生活70
使命价值60
薪资福利
50较低
The job does not disclose salary, and being an intern in Australia may offer moderate compensation. Benefits are not explicitly mentioned.
薪资信号未披露（AI估算：15K-25K/月）
成长发展
90较高
The role offers significant growth through hands-on work with cutting-edge 3D vision and generative AI, plus potential for publishing papers.
技术前沿前沿/新兴技术
技术栈3D Reconstruction、Diffusion Models、PyTorch、SMPL、AIGC
成长机会research papers、real product deployment
业务类型profit_center
工作生活
70中等
The position supports remote work, providing flexibility. However, location in Sydney and potential time zone differences may be challenges.
工作模式远程工作
办公地点海外（不适用）
加班情况未提及（无法判断）
使命价值
60中等
The role has social impact in enhancing user experiences through beauty technology, but the core purpose is commercial.
行业发展高速增长赛道
社会影响中性/一般
使命信号使创作过程更加简单高效
创新程度积极采用新技术
Watch Jobs