Overview
Who this competition fits
Advanced practitioners with experience in general machine learning. Requires strong technical foundations and comfort with complex ML pipelines.
Read the original official blurb
FaceChain是一个以人像为核心的文生图框架。用户仅需要提供最低一张人像照片，结合不同的风格模型，可以生成极具创意的个人AI写真作品。随着FaceChain应用领域的不断拓展，用户对个人AI写真的生成速度和质量都提出了更高的要求。通过本项赛事，我们希望参赛者在提高AI写真的人像逼真度，以及提升风格模型的生成效果、拓展AI人像视频功能上提出优秀的解决方案，共同打造一款高效、真实、风格优美的AI人像写真、视频生成框架。基本组件可包括： 1. ControlNet：人脸、与人体的关键点作为生成图片的参考。 2. Face Adapter（FACT）：用于送入人脸的表征至生成模型中。 3. LoRA：用于控制生成的风格。 注意：由于本赛题仅考察的是生成能力，设计的模型与框架中不应包含换脸模块。 Official category: 实战竞技赛. Organizers: 杭州阿里巴巴飞天信息技术有限公司. Participants: 1000.
Preparation
From registration to a first submission
- 01
Python
- 02
scikit-learn or similar ML library
- 03
data analysis with pandas
Before you commit: Requires solid machine learning fundamentals and good experimental methodology. The competitive landscape is strong, with many experienced teams.
Source
How this page was assembled
Competition information is structured from the official page. Scores are platform estimates for decision support; official rules take precedence.
- Official competition page
- OpenAtom
- Last checked
- Jan 31, 2024