Checklist OPEIA 2025 Competition · Task 3
AI Challenge: Mission EMO — Teaching Machines to Feel
Train a CNN to classify facial expressions from 48×48 grayscale FER2013 images into seven emotions.
The task
Contestants implement a convolutional neural network that maps a 48×48 grayscale face to one of seven emotions (angry, disgust, fear, happy, sad, surprise, neutral), using the FER2013 dataset organised in train/ and test/ class folders, and report the loss, balanced accuracy and F1-score on the validation set with attention to class imbalance.
Abridged by SOTA from the official materials. The official statement has the exact rules, and it wins wherever this summary differs.
At a glance
- You get
- FER2013 images (ZIP downloaded from Google Drive in the notebook).
- You submit
- The completed notebook showing loss, balanced accuracy and F1-score.
- Scoring
- Balanced accuracy and F1-score (as reported by the contestant).
- Format
- OPEIA 2025, a 14-day team competition (teams of four listed on the results page); dates not stated on the pages opened.