cassiekang/cub200_dataset

---
license: mit
task_categories:
- image-classification
language:
- en
tags:
- biology
- birds
- fine-grained image classification
- natural language description
size_categories:
- 1K<n<10K
---
# Dataset Card for CUB_200_2011

## Dataset Description

- **Homepage:**
https://www.vision.caltech.edu/datasets/cub_200_2011/
- **Citation:**
@techreport{WahCUB_200_2011,
Title = ,
Author = {Wah, C. and Branson, S. and Welinder, P. and Perona, P. and Belongie, S.},
Year = {2011}
Institution = {California Institute of Technology},
Number = {CNS-TR-2011-001}
}

### Dataset Summary

The Caltech-UCSD Birds 200-2011 dataset (CUB-200-2011) is an extended version of the original CUB-200 dataset, featuring photos of 200 bird species primarily from North America. This 2011 version significantly expands its predecessor by doubling the number of images per class and introducing new part location annotations, alongside collecting detailed natural language descriptions for each image through Amazon Mechanical Turk (AMT). The dataset includes a total of 11,788 images, split into 5,994 for training and 5,794 for testing.

### Supported Tasks and Leaderboards

This dataset can support a variety of computer vision tasks, including but not limited to:

* Fine-Grained Image Classification
* Object Detection and Localization
* Semantic Segmentation
* Attribute-Based Recognition
* Multitask Learning

### Languages

The dataset includes annotations in English

## Dataset Structure

### Data Instances


A data instance in the CUB-200-2011 dataset comprises an image of a bird species, along with annotations including bounding boxes, part locations, binary attributes, and natural language descriptions.
```
{
"text": "A photo of a Tropical King Bird",
"image": cassiekang/train-00000-of-00001-246c29c8515f0b3f/Tropical_Kingbird_0064_69889.jpg
}
```

### Data Fields

* images: Photographs of birds across 200 species.
* annotations: This includes:
* bounding boxes: Specify the bird's location within the image.
* segmentation labels: Provide pixel-wise segmentation for precise object segmentation.
* part locations: 15 specific parts of the bird are annotated for detailed analysis.
* binary attributes: 312 attributes indicating the presence or absence of certain features or behaviors.
* natural language descriptions: Ten single-sentence descriptions per image, collected via AMT.

### Data Splits

* Training set: 5,994 images
* Test set: 5,794 images

## Considerations for Using the Data

### Social Impact of Dataset

The dataset contributes to advancements in computer vision, particularly in fine-grained image classification and object detection, with potential applications in biodiversity monitoring and species conservation.

原始信息汇总

数据集概述

数据集描述

• 名称: Caltech-UCSD Birds 200-2011 (CUB-200-2011)
• 类别:
  • 任务: 图像分类
  • 语言: 英语
  • 标签: 生物学, 鸟类, 细粒度图像分类, 自然语言描述
• 大小: 1K<n<10K
• 概述: 该数据集包含200种北美鸟类的照片，共计11,788张图像，分为5,994张训练图像和5,794张测试图像。数据集不仅提供了图像，还包括边界框、部分位置、二进制属性和自然语言描述。

数据集结构

数据实例

每个数据实例包括一张鸟类图像及其相关注释，如边界框、部分位置、二进制属性和自然语言描述。

数据字段

• 图像: 200种鸟类的照片。
• 注释:
  • 边界框: 指定鸟在图像中的位置。
  • 分割标签: 提供像素级的精确对象分割。
  • 部分位置: 对鸟的15个特定部分进行注释。
  • 二进制属性: 312个属性，指示特定特征或行为的存在与否。
  • 自然语言描述: 每张图像有10个单句描述，通过AMT收集。

数据分割

• 训练集: 5,994张图像
• 测试集: 5,794张图像

数据集应用

该数据集支持多种计算机视觉任务，包括细粒度图像分类、对象检测和定位、语义分割、属性基识别和多任务学习。