FER2025: Explainable Gender-Aware Facial Emotion Detection via Hybrid Deep Learning and Customized Visual Dataset
Authors/Creators
- 1. Student
- 2. Bsc. in Dept. of CSE, United International University (UIU)
Description
This dataset was created for a deep learning project on facial emotion recognition with gender classification. It contains 7,386 images and 11,253 annotations sourced from copyright-free platforms, ensuring legal and unrestricted use. The dataset includes 12 classes representing six emotions (happy, sad, angry, fear, surprised, sleepy), each labeled for male and female subjects. It is split into training (70%), validation (20%), and test (10%) sets. The metadata_FER2025.csv file provides structured details for each image, including set type, class information, and bounding box coordinates, enabling efficient object detection model training.
Methods
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 deep learning models on the dataset and evaluate its performance using the validation and test sets.
Files
Additional details
Dates
- Available
-
2026-04-21
Software
- Repository URL
- https://github.com/sadman-adib/Explainable-Gender-Aware-Facial-Emotion-Recognition.git
- Programming language
- Python
- Development Status
- Active