Overview
Who this competition fits
适合已有 PaddlePaddle 训练经验,并能处理视觉模型架构表示与多任务预测的开发者。参赛者需要以模型结构编码为输入,预测人脸、人体、车辆和商品等视觉任务上的模型排序;比赛仅支持个人参赛。
Read the original official blurb
从垂类应用出发,选择了人脸、人体、车辆、商品四个任务来训练视觉模型大一统模型。 Official tags: CV,AutoDL. Sign-ups: 1760.
Resources
What this competition gives you
There is a baseline you can run as-is — starting from it is usually far faster than building from scratch.
Preparation
From registration to a first submission
- 01
PaddlePaddle 2.0+
- 02
Vision Transformer 架构编码
- 03
多任务学习
- 04
学习排序
- 05
Kendall tau 优化
- 06
JSON 文件处理
Before you commit: 核心难点不是直接训练视觉识别模型,而是从有限的已评估子网样本中学习架构—性能关系,并为大量未见结构同时给出多个任务的整数排序。提交还必须严格按 JSON 格式填写各任务 rank,且最终排名以隐藏的 B 榜为准。
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
- Baidu AI Studio
- Last checked
- Not recorded