Improving the quality of Indian celebrity data for machine learning applications
Jyoti Kesarwani, Himanshu Rai · Computational Methods in Science and Technology · 2024
This paper presents a comparative study of various convolutional neural network (CNN) models like Inception, ResNet, and VGG16 for face recognition on the Indian Celebrity Dataset. The work highlights the importance of an Indian celebrity face dataset to improve facial recognition performance globally, given the massive and multicultural population in India. The paper discusses different data preprocessing techniques applied to the Indian Celebrity Dataset including data cleaning, reduction, and transformation methods like normalization. It investigates how these preprocessing steps can enhance the data quality and model performance. The CNN models are evaluated on tasks such as facial attribute prediction and face recognition using the preprocessed Indian Celebrity Dataset. The paper aims to provide insights into effectively utilizing CNNs and data preprocessing for face analysis on a diverse Indian celebrity face dataset.