Date of Publication :5th February 2025

Abstract: The aim of comparative study of deep learning algorithms and their frameworks is to interpret women violence and their recent need to monitor and develop the Artificial intelligence based predictive and preventive model that contributes to the issues occurring with Women’s in different forms of violence, especially this research work predict the Human Trafficking, Emotional Abuse, Gender Based violence, Sexual Harassment, Intimate Partner violence, Child molestation and Harassment, Brutality and Domestic violence etc.as well as the comparative performance analysis of deep learning frameworks. The research focuses on Violence against women which effectively used tensor flow and EfficientNet techniques. These Deep Learning framework for predicting Women Violence incorporates five pre trained frameworks i.e. Residual Neural Network50 (ResNet50), Inception v3, and Visual Geometry Group (Vgg16). ResNet50 has 50 layers, Inception v3 has 48 layers and Vgg16 has 16 weighted layers, EfficientNet, MobileNetV2. The Deep learning classifiers that are used for tag based image processing and enhance the quality output for prediction of task. Tag based image Dataset consist of nine types of women violence with 944 images. The deep learning frameworks shows the result based on provided trained dataset to the pre trained frameworks. The experimental results show ResNet50 got accuracy with 72 %, Vgg16 got accuracy with 84 % and InceptionV3 got accuracy with 79 %. The highest accuracy compared with all other frameworks is of Vgg16 framework with 84 %. CNN EfficientNet with 69 % MobileNetV2 with 74 %. The hidden nature of such violence and its frequent underreporting make it a critical area for research. Recent developments in artificial intelligence offer new avenues for detecting and predicting instances of women violence through Deep Learning algorithms.

Reference :