多模态大模型训练数据以及微调数据格式

多模态数据，尤其是中文多模态数据，找一些中文多模态的数据

中文多模态数据集汇总_数据集-阿里云天池本文整理汇总了业界常用的多模态中文数据集，提供了每个数据集的简介、官网、下载地址、Github代码等信息，方便算法研究人员学习研究。https://tianchi.aliyun.com/dataset/145784

LMM 视觉描述（Captioning）和定位（Grounding）数据集本文介绍了 Visual Captioning 和 Visual Grounding 的相关概念、细分领域，然后具体介绍了相关的常见评估集，比如 COCO Caption、NoCaps、Flickr30KRefCOCO等。https://mp.weixin.qq.com/s?__biz=Mzk0ODU3MjcxNA==&mid=2247484571&idx=1&sn=0105194bd7493c33fd8d53be97f688b7&chksm=c364c3def4134ac8512d1d64c7e789e1edabe48e347652c73dc32492cec5627fbdb88cd2eaff&scene=21#wechat_redirect

LMM 视觉问答（VQA）数据集解读本文整理了常用的 VQA 评估集，分为两类，一类是通用型 VQA（General VQA），也就是直接基于图像来问题；另一类是文本导向的 VQA（Text-Oriented VQA），以便了解各个评估集的具体内容和相应的应用场景。https://mp.weixin.qq.com/s?__biz=Mzk0ODU3MjcxNA==&mid=2247484515&idx=1&sn=94b77ab783e903632c83a952a85a4c3c&chksm=c364c326f4134a30c59e9fe1a5a93e030cebb5a893108677d3209e1c17a06c7b7bf4e16d8c51&scene=21#wechat_redirect

GitHub - BradyFU/Awesome-Multimodal-Large-Language-Models: :sparkles::sparkles:Latest Papers and Datasets on Multimodal Large Language Models, and Their Evaluation.:sparkles::sparkles:Latest Papers and Datasets on Multimodal Large Language Models, and Their Evaluation. - BradyFU/Awesome-Multimodal-Large-Language-Modelshttps://github.com/BradyFU/Awesome-Multimodal-Large-Language-Models?tab=readme-ov-file#awesome-datasets

1.XrayGLM

https://github.com/WangRongsheng/XrayGLMhttps://github.com/WangRongsheng/XrayGLM6423张Xray图片，

2.sharegpt4v

用的gpt4v标注的多模态数据，1.2M

InternLM-XComposer/projects/ShareGPT4V/docs/Data.md at main · InternLM/InternLM-XComposer · GitHubInternLM-XComposer2 is a groundbreaking vision-language large model (VLLM) excelling in free-form text-image composition and comprehension. - InternLM-XComposer/projects/ShareGPT4V/docs/Data.md at main · InternLM/InternLM-XComposerhttps://github.com/InternLM/InternLM-XComposer/blob/main/projects/ShareGPT4V/docs/Data.md3.wukong

4.zero-chinese

5.laion-2b-chinese

数据格式：

这里llava格式是应用最为广泛的，除了qwen-vl系列，几乎所有的系列都是llava格式。

1.InternLM-VL

InternLM-XComposer/finetune/README_zh-CN.md at main · InternLM/InternLM-XComposer · GitHubInternLM-XComposer2 is a groundbreaking vision-language large model (VLLM) excelling in free-form text-image composition and comprehension. - InternLM-XComposer/finetune/README_zh-CN.md at main · InternLM/InternLM-XComposerhttps://github.com/InternLM/InternLM-XComposer/blob/main/finetune/README_zh-CN.md2.QWEN-VL

Qwen-VL/README_CN.md at master · QwenLM/Qwen-VL · GitHubThe official repo of Qwen-VL (通义千问-VL) chat & pretrained large vision language model proposed by Alibaba Cloud. - Qwen-VL/README_CN.md at master · QwenLM/Qwen-VLhttps://github.com/QwenLM/Qwen-VL/blob/master/README_CN.md

python 复制代码

[
  {
    "id": "identity_0",
    "conversations": [
      {
        "from": "user",
        "value": "你好"
      },
      {
        "from": "assistant",
        "value": "我是Qwen-VL,一个支持视觉输入的大模型。"
      }
    ]
  },
  {
    "id": "identity_1",
    "conversations": [
      {
        "from": "user",
        "value": "Picture 1: <img>https://qianwen-res.oss-cn-beijing.aliyuncs.com/Qwen-VL/assets/demo.jpeg</img>\n图中的狗是什么品种？"
      },
      {
        "from": "assistant",
        "value": "图中是一只拉布拉多犬。"
      },
      {
        "from": "user",
        "value": "框出图中的格子衬衫"
      },
      {
        "from": "assistant",
        "value": "<ref>格子衬衫</ref><box>(588,499),(725,789)</box>"
      }
    ]
  },
  { 
    "id": "identity_2",
    "conversations": [
      {
        "from": "user",
        "value": "Picture 1: <img>assets/mm_tutorial/Chongqing.jpeg</img>\nPicture 2: <img>assets/mm_tutorial/Beijing.jpeg</img>\n图中都是哪"
      },
      {
        "from": "assistant",
        "value": "第一张图片是重庆的城市天际线，第二张图片是北京的天际线。"
      }
    ]
  }
]

3. swift中的数据格式

qwen：

python 复制代码

[
    {"conversations": [
        {"from": "user", "value": "Picture 1:<img>img_path</img>\n11111"},
        {"from": "assistant", "value": "22222"}
    ]},
    {"conversations": [
        {"from": "user", "value": "Picture 1:<img>img_path</img>\nPicture 2:<img>img_path2</img>\nPicture 3:<img>img_path3</img>\naaaaa"},
        {"from": "assistant", "value": "bbbbb"},
        {"from": "user", "value": "Picture 1:<img>img_path</img>\nccccc"},
        {"from": "assistant", "value": "ddddd"}
    ]},
    {"conversations": [
        {"from": "user", "value": "AAAAA"},
        {"from": "assistant", "value": "BBBBB"},
        {"from": "user", "value": "CCCCC"},
        {"from": "assistant", "value": "DDDDD"}
    ]}
]

4.bunny

python 复制代码

{
'id': '0', 
'image': 'coco_2017/000000337760.jpg', 
'conversations': [
    {'from': 'human', 'value': '<image>\n这是一辆现代消防车吗？\n请用单个短语回答问题。'}, 
    {'from': 'gpt', 'value': '不'}, 
    {'from': 'human', 'value': '这张照片是黑白的还是彩色的？'}, 
    {'from': 'gpt', 'value': '黑白的'},  
    {'from': 'human', 'value': '图中显示了哪些车辆？'}, 
    {'from': 'gpt', 'value': '消防车'}
]
}

5.visualglm

python 复制代码

[
    {"img": "fewshot-data/2p.png", "prompt": "这张图片的背景里有什么内容？", "label": "这张图片的背景是蒙蒙细雨。"},
    {"img": "fewshot-data/pig.png", "prompt": "这张图片的背景里有什么内容？", "label": "这张图片的背景是是虚化的。"},
    {"img": "fewshot-data/meme.png", "prompt": "这张图片的背景里有什么内容？", "label": "这张图片的背景是蓝色的木质地板。"},
    {"img": "fewshot-data/passport.png", "prompt": "这张图片的背景里有什么内容？", "label": "这张图片的背景是棕黄色木质桌子。"},
    {"img": "fewshot-data/tower.png", "prompt": "这张图片的背景里有什么内容？", "label": "这张图片的背景是黄昏的天空、云彩和繁华的城市高楼。"},
    {"img": "fewshot-data/rub.png", "prompt": "这张图片的背景里有什么内容？", "label": "这张图片的背景是太阳、大树、蓝天白云。"},
    {"img": "fewshot-data/push.png", "prompt": "这张图片的背景里有什么内容？", "label": "这张图片的背景是蓝天和沙漠。"},
    {"img": "fewshot-data/traf.png", "prompt": "这张图片的背景里有什么内容？", "label": "这张图片的背景是城市街道。"},
    {"img": "fewshot-data/music.png", "prompt": "这张图片的背景里有什么内容？", "label": "这张图片的背景是一个音乐混音器。"},
    {"img": "fewshot-data/pattern.png", "prompt": "这张图片的背景里有什么内容？", "label": "这张图片的背景是小区的楼房和街道。"},
    {"img": "fewshot-data/rou.png", "prompt": "这张图片的背景里有什么内容？", "label": "这张图片的背景是大理石桌子和一个盘子。"},
    {"img": "fewshot-data/katong.png", "prompt": "这张图片的背景里有什么内容？", "label": "这张图片的背景是绿色的草地。"},
    {"img": "fewshot-data/man.jpg", "prompt": "这张图片的背景里有什么内容？", "label": "这张图片的背景是城市的街道和高楼。"},
    {"img": "fewshot-data/kobe.png", "prompt": "这张图片的背景里有什么内容？", "label": "这张图片的背景是虚化的观众席。"},
    {"img": "fewshot-data/panda.jpg", "prompt": "这张图片的背景里有什么内容？", "label": "这张图片的背景是纯白的。"},
    {"img": "fewshot-data/titan.png", "prompt": "这张图片的背景里有什么内容？", "label": "这张图片的背景是一座雕像。"},
    {"img": "fewshot-data/woman.png", "prompt": "这张图片的背景里有什么内容？", "label": "这张图片的背景是纯蓝的。"},
    {"img": "fewshot-data/ghost.jpg", "prompt": "这张图片的背景里有什么内容？", "label": "这张图片的背景是一个房间。"},
    {"img": "fewshot-data/justice.png", "prompt": "这张图片的背景里有什么内容？", "label": "这张图片的背景是天空和阳光。"},
    {"img": "fewshot-data/tianye.png", "prompt": "这张图片的背景里有什么内容？", "label": "这张图片的背景是金黄的田野。"}
]