rgonzalezn/closed-open-eyes

👀 Open and Closed Eyes Dataset

Welcome to the Open and Closed Eyes Dataset! This dataset is designed to help researchers and developers in computer vision and machine learning tasks, particularly in recognizing and distinguishing between open and closed eyes in various contexts. Below, you'll find a detailed description of the dataset structure, categories, and how to interpret the data. 🌟

📁 Dataset Structure

The dataset is stored in Parquet files, ensuring efficient storage and quick access. Each Parquet file contains records of images along with their associated metadata. The files are named in the format dataset_XXX.parquet, where XXX is a sequential number.

Data Elements

Each record in the Parquet files includes the following elements:

• Image_id: A unique identifier for each image.
• Left_eye_react: Bounding box coordinates for the left eye.
• Right_eye_react: Bounding box coordinates for the right eye.
• Label: Indicates whether the eyes are 'open_eyes' or 'closed_eyes'.
• Image_data:
  • file: The image data in byte format.
  • filename: The filename, derived from the SHA256 checksum of the image.

📊 Categories and Attributes

The dataset features a wide range of categories to ensure diverse representation. The images in the dataset are AI-generated and include around 126,000 balanced samples. Here’s a breakdown of the categories:

• 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

🔍 How to Use the Dataset

1. Accessing the Data: Load the Parquet files using libraries like PyArrow or Pandas to easily manipulate the data in Python.
2. Image Data: The Image_data field contains the image bytes. You can use libraries such as PIL or OpenCV to decode and display the images.
3. Bounding Boxes: The bounding box coordinates (Left_eye_react, Right_eye_react) help in precisely locating the eyes within the images.
4. Labels and Metadata: Use the provided labels and metadata to filter and categorize the images as per your requirements.

This dataset is rich in variety and detail, making it an excellent resource for developing robust models and conducting comprehensive analyses. Dive in and explore the diversity of human features and settings captured in this unique collection! 🚀

Happy exploring! 🌈✨

📜 Ethical Considerations and Limitations

• Potential NSFW Content: While we have taken steps to exclude inappropriate content, there may still be some images that are not safe for work (NSFW). Users should exercise caution when browsing the dataset.
• Cultural Specificity: This dataset includes a variety of cultural contexts and settings, but it may not fully represent all cultural nuances. Users should be aware of the potential for cultural bias in the imagery and scenarios presented.
• Potential Biases: There may be biases inherent in the generation process, including those related to gender, age, and ethnicity. It's important for users to be mindful of these biases when using the dataset for training and evaluation.
• Image Diversity and Complexity: While the dataset aims to cover a wide range of scenarios and attributes, some aspects of image diversity and complexity might still be limited.

🔧 Dataset Maintenance

If you find any mistakes or errors, please start a new discussion or open a pull request on our repository. Your contributions are invaluable in improving the dataset and ensuring its accuracy.

📚 Citation

If you use the Open and Closed Eyes Dataset in your research, please cite it as:

@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},
}

Thank you for your interest in our dataset! We hope it proves to be a valuable resource for your research and development projects.