FER2025: A Deep Learning Approach to Facial Emotion Recognition with Gender Classification Using CNN
Description
This dataset was developed for a deep learning project focused on facial emotion recognition with gender classification using Convolutional Neural Networks (CNNs). It comprises a total of 3,176 images and 5,160 annotations, collected exclusively from reputable sources offering copyright-free or royalty-free images, such as Unsplash, iStock, Shutterstock, Getty Images etc. These platforms explicitly permit usage for personal and commercial purposes without requiring attribution. The dataset includes 12 distinct classes, corresponding to six emotion categories (e.g., happy, sad, angry, fear, surprised, sleepy), each labeled separately for male and female subjects, resulting in a total of 12 unique class labels. To facilitate effective model training and evaluation, the dataset is organized into three subsets: Training Set: 70% of the data Validation Set: 20% of the data Test Set: 10% of the data The metadata_FER2025.csv file lists each image of a facial emotion with its name, set type (Train/Valid/Test), Class ID, Class Name, and bounding box details (X, Y, Width, Height). It organizes the dataset for efficient object detection model training. All annotations were carefully curated to ensure compliance with licensing terms and legal usage rights.
Files
Steps to reproduce
To reproduce the dataset, begin by collecting copyright-free images from platforms like Unsplash, iStock, Shutterstock, Getty Images etc. Annotate each image with one of 12 class labels representing six emotions for both male and female subjects. Organize the dataset into three directories: 70% for training, 20% for validation, and 10% for testing. Preprocess the images through resizing, normalization, and optional augmentation. Finally, train CNN models on the dataset and evaluate its performance using the validation and test sets.
Institutions
- United International University