小米3D/4D重建生成算法实习生
任职要求
- 目前正在攻读计算机科学、电子工程、机器人学或相关领域的硕士或博士学位。 - 具备扎实的 C++ 和 Python 编程能力。 - 熟悉 PyTorch 或 TensorFlow 等深度学习框架；对神经重建技术（如 3DGS、NeRF）有基本理解或相关经验。 - 具备 3DGS 重…
工作职责
- 参与跨团队协作，将研究原型转化为可扩展的系统。 - 参与神经重建流水线中算法与模型的开发。 - 探索并实验前沿（SOTA）神经重建方法，例如 3DGS、NeRF、Gaussian Splatting 等。 - 支持模型训练、评估与集成，服务于自动驾驶真实应用场景。 - 构建并维护用于评估和可视化模型性能的工具。 - 产出高质量论文，或将算法应用于实际业务场景中。
• Collaborating with a top-tier team to improve and refine 3D and 4D reconstruction algorithms. • Engaging in the refinement of computer vision models, with a specific emphasis on 3D GS reconstruction and integrating feed-forward models. • Conducting research and experiments to push the boundaries of computer vision and AI. • Analyzing and interpreting complex data to determine effective solutions and improvements. • Using the 3D reconstruction results to accelerate AV (autonomous vehicle) simulation and development. • Sharing your findings and contributions with both technical and non-technical collaborators, helping to build the strategic direction of our projects.
We are looking for a passionate team member to contribute to our autonomous driving product! The role involves applying deep learning and computer vision technologies to 3D/4D world modeling for autonomous vehicles. Our work is crucial to empower the training and testing of end-to-end driving models through end-to-end close-loop simulation. You will have the opportunity to work with a diverse team of researchers and engineers in the field of 3D/4D reconstruction, world modeling and simulation to deliver impact to our customers around the world! What you’ll be doing: • Build innovative world reconstruction systems using large geometry models, perception models, Gaussian splatting, diffusion models, and generative world models. • Invent evaluation methods to measure the reconstruction and simulation quality. • Develop tools to visualize and triage evaluation metrics. • Build automated and agentic workflows to boost developer efficiency. • Profile and optimize the performance of neural networks training and deployment. • Relentlessly improve the reconstruction fidelity and simulation realism. • Collaborate with scientists and developers across several organizations.
1.负责项目中的场景原画的整体设计，单体建筑的设计拆分； 2.跟进美术最终表现效果，提出改善意见并实施修改； 3.跟进原画外包，制定外包规范； 4.跟进3D场景引擎内的效果，配合3D场景完善场景制作。
1.独立负责偏写实风格怪物战斗动画的设计与制作； 2.配合其他部门，保证战斗动作的写实度和设计个性，最终让怪物表现在游戏内达到3A品质； 3.梳理与指导角色动画设计规范与方向（包括不限于生产管线，动捕管线，指导外包等）。