MichalMlodawski/closed-open-eyes|图像分类数据集|目标检测数据集

Open and Closed Eyes数据集是一个用于计算机视觉和机器学习任务的数据集，特别适用于识别和区分不同场景下的睁眼和闭眼状态。数据集包含约126,000个平衡样本，存储为Parquet文件，每个记录包含图像ID、左右眼的边界框坐标、标签（睁眼或闭眼）以及图像数据。数据集涵盖了多种类别，包括年龄、性别、场景、肤色、面部焦点、服装、发型、时间、天气、情绪和配饰等。此外，README还提供了数据集的伦理考虑、维护建议和引用方式。

The Open and Closed Eyes Dataset is designed for computer vision and machine learning tasks, particularly for recognizing and distinguishing between open and closed eyes. This dataset includes AI-generated images with a balanced distribution across various categories such as age groups, gender, scenery, skin color, face focus, clothing, hairstyle, time of day, weather, emotion, and accessories. The dataset is stored in Parquet files and includes metadata like image ID, bounding box coordinates for eyes, labels indicating open or closed eyes, and image data including the file and filename. The dataset is intended to be used for training and evaluating models, with ethical considerations and potential biases noted.

原始信息汇总

Open and Closed Eyes Dataset

概述

• 名称: Open and Closed Eyes Dataset
• 语言: 英语
• 大小: 100K<n<1M
• 任务类别: 图像分类, 目标检测
• 标签: ai-generated, balanced-dataset
• 许可证: odc-by

数据集结构

• 存储格式: Parquet文件
• 文件命名: dataset_XXX.parquet

数据元素

• Image_id: 图像的唯一标识符
• Left_eye_react: 左眼边界框坐标
• Right_eye_react: 右眼边界框坐标
• Label: 眼睛状态标签（open_eyes 或 closed_eyes）
• Image_data:
  • file: 图像数据（字节格式）
  • filename: 图像文件名（基于SHA256校验和）

类别和属性

• Eyes:
  • Open eyes
  • Closed eyes
• Age Groups:
  • Infant, Young adult, Adult, Middle-aged, Senior, Elderly
• Gender:
  • Female, Male
• Scenery:
  • Inside train, Beach, Mountain, City, Forest, Desert, Car interior, Kitchen, Park, Office, Home living room, Space, Underwater, Airport terminal, Concert hall, Museum, Gym, Restaurant, Library, Farm, Art gallery, Rooftop, Garden, Cave, Waterfall, Castle interior, Shopping mall, University lecture hall, Ski resort lodge, Tropical beach house, Ancient temple interior, Futuristic city apartment, Hot air balloon basket, Carnival tent, Haunted house interior, Volcano observatory, Space station interior, Jungle treehouse, Arctic research station, Savanna safari camp, Cozy bedroom, Modern bathroom, Stylish home office, Luxurious hotel room, Rustic cabin interior, Industrial loft, Minimalist studio apartment, Vintage diner, School classroom, Hospital ward, Movie theater, Computer server room, Greenhouse interior, Subway station, Airport control tower, Lighthouse interior, Medieval tavern, Futuristic laboratory, Underground bunker, Treehouse interior, Ancient ruins, Underwater cave, Zen garden, Post-apocalyptic cityscape, Steampunk workshop, Fairy tale cottage, Cyberpunk street, Floating sky island, Abandoned amusement park, Crystal cave, Alien planet landscape, Medieval castle courtyard, Deep space nebula, Rainforest canopy, Arctic ice shelf, Volcanic landscape, Bustling bazaar, Tranquil monastery, Neon-lit nightclub, Retro 1950s diner
• Skin Color:
  • White, Black, Brown, Light, Dark, Olive, Tan, Albino, Freckled, Vitiligo, Reddish, Yellowish
• Face Focus:
  • Focus on left side, Focus on right side, Focus on center, Focus on top, Focus on bottom, Focus on eyes, Focus on nose, Focus on mouth, Focus on chin, Focus on forehead, Full face focus
• Clothing:
  • Casual, Formal, Sports, Traditional, Futuristic, Summer, Winter, Business, Swimwear, Nightwear, Costume, Uniform, Vintage, Bohemian, Punk, Gothic, High fashion, Streetwear, Cyberpunk, Steampunk, Medieval, Renaissance, Space suit, Superhero costume, Military uniform, Royalty attire, Hippie, Grunge, Preppy, Hip-hop fashion, Emo fashion
• Hairstyle:
  • Short hair, Long hair, Curly hair, Straight hair, Braided hair, Bald, Wavy hair, Ponytail, Buzz cut, Dreadlocks, Mohawk, Afro, Pixie cut, Bob cut, Undercut, Mullet, Side-swept hair, Spiky hair, Slicked back hair, Messy hair, Ombre hair, Highlighted hair, Two-toned hair, Asymmetrical hair, Pompadour, Quiff, Faux hawk, Bowl cut, Shag haircut, Layered hair, Feathered hair, Cornrows, Man bun, Topknot, Crown braid, Fishtail braid
• Time of Day:
  • Early morning, Mid-morning, Late morning, Noon, Early afternoon, Mid-afternoon, Late afternoon, Early evening, Dusk, Night, Midnight, Pre-dawn
• Weather:
  • Sunny, Partly cloudy, Overcast, Light rain, Heavy rain, Thunderstorm, Snowy, Blizzard, Foggy, Misty, Windy, Calm, Hail, Sleet, Hurricane, Tornado, Sandstorm, Heat wave, Cold snap, Rainbow
• Emotion:
  • Happy, Sad, Angry, Surprised, Neutral, Scared, Disgusted, Confused, Excited, Thoughtful, Amused, Bored, Confident, Curious, Embarrassed, Proud, Relieved, Anxious, Hopeful, Determined
• Accessories:
  • Hat, Cap, Beanie, Scarf, Earrings, Necklace, Bracelet, Ring, Watch, Tie, Bow tie, Bandana, Headband, Hair clip, Belt, Suspenders, Gloves, Handbag, Backpack, Umbrella, Cane, Walking stick, Monocle, Pocket watch, Brooch, Lapel pin, None

使用方法

1. 访问数据: 使用PyArrow或Pandas加载Parquet文件。
2. 图像数据: 使用PIL或OpenCV解码和显示图像。
3. 边界框: 使用边界框坐标精确定位眼睛。
4. 标签和元数据: 根据提供的标签和元数据过滤和分类图像。

伦理考虑和限制

• 潜在的NSFW内容: 可能包含不适合工作环境的内容。
• 文化特异性: 可能不完全代表所有文化背景。
• 潜在偏见: 生成过程中可能存在性别、年龄和种族偏见。
• 图像多样性和复杂性: 某些方面的多样性和复杂性可能有限。

引用

@misc{open_closed_eyes2024, author = {Michał Młodawski}, title = {Open and Closed Eyes Dataset}, month = July, year = 2024, url = {https://huggingface.co/datasets/MichalMlodawski/closed-open-eyes}, }