220k-GPT4Vision-captions-from-LIVIS

# 220k-GPT4Vision-captions-from-LVIS
## by: Christoph Schuhmann, Peter Bevan, 21 Nov, 2023

---

This dataset comprises 220,000 captioned images from the LVIS dataset. The captions were generated by summarising the [LVIS-Instruct4V](https://huggingface.co/datasets/X2FD/LVIS-Instruct4V) dataset released by X2FD. The instructions are converted into captions using [Mistral-7B-OpenOrca](https://huggingface.co/Open-Orca/Mistral-7B-OpenOrca).

---

### PROMPT
`"""<<SYS>> You are a highly intelligent, empathic, helpful, respectful, and honest assistant with high emotional intelligence.
Always answer as helpfully and honest as possible, while being safe. Your answers should not include any harmful, unethical, racist,
sexist, toxic, dangerous, or illegal content. Please ensure that your responses are socially unbiased and positive in nature.
If a question does not make any sense, or is not factually coherent, explain why instead of answering something not correct.
If you don't know the answer to a question, please don't share false information. <</SYS>> DIALOGUE: {text} INSTRUCTIONS:
The previous DIALOGUE is a conversation between a chatbot and a user about an image. Please summarize all information and details about
the image the chatbot is talking about in DIALOGUE in one precise, very factual caption with as many details as you can extract from DIALOGUE.
Do not make up details about the image and stick strickly to the information in DIALOGUE. Only include factual, descriptive details about the image.
Start with the words "This image showcases":"""`

"This image showcases" was trimmed from the beginning of each caption upon generation.

---

# Citation
```bibtex
@misc{LAION_LVIS_220,
title = {220k-GPT4Vision-captions-from-LVIS},
author = {Christoph Schuhmann and Peter Bevan},
year = {2023},
publisher = {HuggingFace},
journal = {HuggingFace repository},
howpublished = {\url{https://huggingface.co/datasets/laion/220k-GPT4Vision-captions-from-LIVIS}},
}
```

# 基于LVIS的22万条GPT4Vision图像描述数据集
## 作者：克里斯托夫·舒曼（Christoph Schuhmann）、彼得·贝文（Peter Bevan），2023年11月21日

---

本数据集包含源自LVIS数据集的22万张配有图像描述的图像样本。其图像描述通过汇总X2FD发布的[LVIS-Instruct4V](https://huggingface.co/datasets/X2FD/LVIS-Instruct4V)数据集生成，并借助[Mistral-7B-OpenOrca](https://huggingface.co/Open-Orca/Mistral-7B-OpenOrca)模型将原数据集的指令转换为最终的图像描述。

---

### 提示词
`"""<<SYS>> 你是一位具备高智能、共情能力、乐于助人、尊重他人且诚实可靠的助手，拥有出色的情绪智力。
请始终在保证安全的前提下，尽可能提供有益且诚实的回答。你的回答不应包含任何有害、不道德、种族主义、性别歧视、具有攻击性、危险或非法的内容，请确保回复符合社会公序良俗且基调积极向上。
若问题逻辑不通或与事实不符，请解释原因而非给出错误答案。若你不清楚某一问题的答案，请不要编造虚假信息。<</SYS>> 对话：{text} 指令：
前述对话为聊天机器人与用户围绕某张图像展开的交流，请从对话中提取所有与该图像相关的信息与细节，将其整合为一段精准、事实性强且尽可能详尽的图像描述。不得凭空捏造图像细节，严格遵循对话中的既有信息，仅保留关于图像的客观描述性细节。请以"This image showcases"作为描述的开头语句："""`

在生成阶段，每条图像描述的开头均被移除了"This image showcases"前缀。

---

# 引用
bibtex
@misc{LAION_LVIS_220,
title = {220k-GPT4Vision-captions-from-LVIS},
author = {Christoph Schuhmann and Peter Bevan},
year = {2023},
publisher = {HuggingFace},
journal = {HuggingFace repository},
howpublished = {url{https://huggingface.co/datasets/laion/220k-GPT4Vision-captions-from-LIVIS}},
}