FER2013

Overview

FER2013 (Facial Expression Recognition 2013) contains 48×48 grayscale face crops, each labelled with one of seven facial expressions: angry, disgust, fear, happy, sad, surprise and neutral. It was released for the ICML 2013 “Challenges in Representation Learning” Kaggle competition and is one of the datasets of CLIP’s linear-probe evaluation suite. The faces were found with web image searches and cropped automatically, so pose, lighting, age and occlusion vary widely; human accuracy is estimated at about 65%, and the classes are imbalanced (disgust has 547 images, happy 8,989).

The competition’s Usage column defines the splits:

Train: 28,709 images (Training)
Validation: 3,589 images (PublicTest)
Test: 3,589 images (PrivateTest)

Download

Kaggle requires an account and acceptance of the competition rules, so FER2013 cannot be downloaded automatically. Download fer2013.csv (or fer2013.tar.gz / icml_face_data.csv) from the competition data page and place it in the download directory (download_dir, by default ~/.stable-datasets/downloads/, or $STABLE_DATASETS_CACHE_DIR/downloads/). The builder verifies the file’s MD5 checksum before using it.

Data Structure

When accessing an example using ds[i], you will receive a dictionary with the following keys:

Key	Type	Description
`image`	`PIL.Image.Image`	48×48 grayscale face
`label`	int	0: angry, 1: disgust, 2: fear, 3: happy, 4: sad, 5: surprise, 6: neutral

Usage Example

from stable_datasets.images import FER2013

# Requires fer2013.csv in the download directory (see "Download" above).
ds = FER2013(split="train")
ds_all = FER2013(split=None)  # train / validation / test

sample = ds[0]
print(sample.keys())  # {"image", "label"}

ds_torch = ds.with_format("torch")

References

Official website: https://www.kaggle.com/c/challenges-in-representation-learning-facial-expression-recognition-challenge
License: Kaggle competition rules

Citation

@inproceedings{goodfellow2013challenges,
    title={Challenges in Representation Learning: A report on three machine learning contests},
    author={Goodfellow, Ian J. and Erhan, Dumitru and Carrier, Pierre Luc and Courville, Aaron and
            Mirza, Mehdi and Hamner, Ben and Cukierski, Will and Tang, Yichuan and Thaler, David and
            Lee, Dong-Hyun and others},
    booktitle={International Conference on Neural Information Processing (ICONIP)},
    year={2013}}