StephanAkkerman/fintwit-images

---
language:
- en
license: mit
task_categories:
- image-classification
- image-feature-extraction
pretty_name: FinTwit Images
dataset_info:
features:
- name: image
dtype: image
- name: label
dtype:
class_label:
names:
'0': charts
'1': non-charts
- name: id
dtype: string
splits:
- name: train
num_bytes: 320924227.116
num_examples: 4177
download_size: 444164916
dataset_size: 320924227.116
configs:
- config_name: default
data_files:
- split: train
path: data/train-*
tags:
- fintwit
- twitter
- charts
- financial
- financial charts
- finance
- stocks
- crypto
- image
---

## FinTwit Images

This dataset is a collection of a sample of images from tweets that I scraped using my [Discord bot](https://github.com/StephanAkkerman/fintwit-bot) that keeps track of financial influencers on Twitter.
The data consists of images that were part of tweets that did not mention a ticker.
This dataset can be used for a wide variety of tasks, such as image classification or feature extraction.

### FinTwit Charts Collection
This dataset is part of a larger collection of datasets, scraped from Twitter and labeled by a human (me). Below is the list of related datasets.
- [Crypto Charts](huggingface.co/datasets/StephanAkkerman/crypto-charts): Images of financial charts of cryptocurrencies
- [Stock Charts](https://huggingface.co/datasets/StephanAkkerman/stock-charts): Images of financial charts of stocks
- [FinTwit Images](https://huggingface.co/datasets/StephanAkkerman/fintwit-images): Images that had no clear description, this contains a lot of non-chart images

## Dataset Structure
Each images in the dataset is structured as follows:

- **Image**: The image of the tweet, this can be of varying dimensions.
- **Label**: A numerical label indicating the category of the image, with '1' for charts, and '0' for non-charts.

## Dataset Size
The dataset comprises 4,579 images in total, categorized into:

- 1,083 chart images
- 3,496 non-chart images

## Usage
I used this dataset for training my [chart-recognizer model](https://huggingface.co/StephanAkkerman/chart-recognizer) for classifying if an image is a chart or not.

## Acknowledgments
We extend our heartfelt gratitude to all the authors of the original tweets.

## License
This dataset is made available under the MIT license, adhering to the licensing terms of the original datasets.

原始信息汇总

FinTwit Images 数据集概述

基本信息

• 语言: 英语
• 许可证: MIT
• 任务类别: 图像分类, 图像特征提取
• 美观名称: FinTwit Images

数据集特征

• 特征:
  • image: 图像数据
  • label: 标签数据，类别包括 charts 和 non-charts
  • id: 字符串类型

数据集分割

• train:
  • 字节数: 320924227.116
  • 样本数: 4177

数据集大小

• 下载大小: 444164916
• 数据集大小: 320924227.116

配置

• default:
  • 数据文件:
    • split: train
    • path: data/train-*

标签

• tags:
  • fintwit
  • twitter
  • charts
  • financial
  • financial charts
  • finance
  • stocks
  • crypto
  • image

数据集结构

• Image: 推文图像，尺寸不一
• Label: 数值标签，1 表示 charts，0 表示 non-charts

数据集大小

• 总图像数: 4579
  • chart 图像数: 1083
  • non-chart 图像数: 3496

使用

• 用于训练 chart-recognizer 模型，用于分类图像是否为 chart。

许可证

• 该数据集在 MIT 许可证下发布。