大规模中文多模态评测基准MUGE

多模态理解和生成评估挑战榜（MUGE）是由阿里巴巴达摩院智能计算实验室发起，由阿里云天池平台承办，并由浙江大学、清华大学等单位共同协办。 MUGE榜单为研究者们提供了统一的数据和评测基准，旨在促进多模态理解与生成领域的技术发展。<br />
官网地址：https://tianchi.aliyun.com/muge <br />
Github：https://github.com/MUGE-2021 , https://github.com/OFA-Sys/Chinese-CLIP <br />

The Multimodal Understanding and Generation Evaluation Challenge Benchmark (MUGE) was initiated by the Intelligent Computing Laboratory of Alibaba DAMO Academy, hosted by the Alibaba Cloud Tianchi Platform, and co-organized by Zhejiang University, Tsinghua University and other institutions. The MUGE Benchmark provides researchers with unified datasets and evaluation benchmarks, aiming to promote technological advancements in the field of multimodal understanding and generation.
Official website: https://tianchi.aliyun.com/muge
Github: https://github.com/MUGE-2021, https://github.com/OFA-Sys/Chinese-CLIP

AI搜集汇总

背景与挑战

背景概述

大规模中文多模态评测基准MUGE是一个专注于中文电商领域的多模态数据集，由阿里巴巴达摩院智能计算实验室发起，旨在为研究者提供统一的数据和评测基准以推动多模态理解与生成技术的发展。该数据集包含三个核心任务：图像描述生成、文本到图像生成和图文检索，覆盖服装、食品、化妆品等多个商品类目，并提供大量图片和文本数据用于模型训练和评估。