ckplus-dataset / README.md
AlirezaF138's picture
Update README.md
059ad99 verified
|
raw
history blame
1.07 kB
metadata
dataset_info:
  features:
    - name: image
      dtype: image
    - name: label
      dtype: string
    - name: file
      dtype: string
  splits:
    - name: train
      num_bytes: 1749995
      num_examples: 981
  download_size: 1845563
  dataset_size: 1749995
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
license: cc0-1.0
task_categories:
  - image-classification
size_categories:
  - n<1K

CK+ Dataset

Overview

The CK+ dataset contains adapted data from the CK+ dataset, which includes 327 images of facial expressions, copied three times for some reason. Each image has been preprocessed to 48x48 pixels, converted to grayscale, and cropped to focus on facial features using haarcascade_frontalface_default. This is a reupload of https://www.kaggle.com/datasets/shawon10/ckplus.

Emotions

The dataset labels emotions based on the following indices:

  • 0: Anger (45 samples)
  • 1: Disgust (59 samples)
  • 2: Fear (25 samples)
  • 3: Happiness (69 samples)
  • 4: Sadness (28 samples)
  • 5: Surprise (83 samples)
  • 7: Contempt (18 samples)