We train 4 CNN architectures for 30 epochs and comapre the classification scores. The data used is taken from UTKFace dataset. Each network achieves 75%+ peak accuracy in classification.
Venkat, N., Srivastava, S. (2018). Ethnicity Detection using Deep Convolutional Neural Networks. DOI: 10.13140/RG.2.2.34591.20642
You can download models and train-test split here
python3 -m pip install -r requirements.txt
Executing the code for training and testing
python3 main.py
Executing the code to view results on the trained models
python3 accuracies.py