Dark Mode
Emotion Recognition Grayscale Images
Data Science and Analytics
Tags and Keywords
Trusted By
"No reviews yet"
Free
About
This dataset is an adapted version of the Cohn-Kanade Dataset (CK+), comprising 920 individual facial expressions. It has been specifically prepared for facial expression recognition tasks, with images reshaped to a 48x48 pixel grayscale format and face-cropped using a Haar classifier. Noisy images, affected by factors like room light, hair format, or skin colour, have been adapted for clearer identification. The dataset columns include emotion, pixels, and usage, categorising expressions into seven distinct emotions plus neutral. It was originally developed for comparison with datasets such as FER2013.
Columns
- emotion: An index representing a specific facial expression.
- 0: Anger (45 samples)
- 1: Disgust (59 samples)
- 2: Fear (25 samples)
- 3: Happiness (69 samples)
- 4: Sadness (28 samples)
- 5: Surprise (83 samples)
- 6: Neutral (593 samples)
- 7: Contempt (18 samples)
- A total of 920 valid samples are present.
- pixels: A string containing 2304 values, representing the 48x48 pixels of each grayscale image. There are 920 unique pixel value strings.
- Usage: Defines the allocation of samples for model training and testing.
- Training: 80% of the dataset.
- PublicTest: 10% of the dataset.
- PrivateTest: 10% of the dataset.
- A total of 920 valid samples are distributed across these categories.
Distribution
The dataset is provided in CSV format (ckextended.csv) and totals 7.58 MB in size. It contains 920 rows or records, with each record representing a single facial expression image. The images are already pre-processed to 48x48 pixels and are in grayscale, ensuring a consistent and ready-to-use format for computer vision applications. Each row's 'pixels' column holds 2304 values, corresponding to the 48x48 pixel dimensions.
Usage
This dataset is ideally suited for:
- Developing and training machine learning models for facial expression recognition.
- Research in computer vision and image processing.
- Comparing the performance of different algorithms for emotion classification.
- Building applications that require automated detection of human emotions.
Coverage
The dataset's scope is primarily focused on facial expressions without specific geographical, time range, or demographic notes provided within the source material.
License
CC0: Public Domain
Who Can Use It
- Researchers in artificial intelligence and computer vision to train and validate emotion recognition models.
- Data scientists and machine learning engineers developing solutions for sentiment analysis or human-computer interaction.
- Academics and students studying facial recognition, image processing, or deep learning.
- Developers creating applications that interpret or react to human emotions.
Dataset Name Suggestions
- Cohn-Kanade Plus Facial Expressions
- Emotion Recognition Grayscale Images
- CK+ Adapted Facial Data
- 48x48 Facial Expression Dataset
Attributes
Original Data Source: Emotion Recognition Grayscale Images
QUALITY