A deep-learning approach to early identification of suggested sexual harassment from videos

Shreya Shetye,Anwita Maiti,Tannistha Maiti,Tarry Singh

2023-06-02

Abstract:Sexual harassment, sexual abuse, and sexual violence are prevalent problems in this day and age. Women's safety is an important issue that needs to be highlighted and addressed. Given this issue, we have studied each of these concerns and the factors that affect it based on images generated from movies. We have classified the three terms (harassment, abuse, and violence) based on the visual attributes present in images depicting these situations. We identified that factors such as facial expression of the victim and perpetrator and unwanted touching had a direct link to identifying the scenes containing sexual harassment, abuse and violence. We also studied and outlined how state-of-the-art explicit content detectors such as Google Cloud Vision API and Clarifai API fail to identify and categorise these images. Based on these definitions and characteristics, we have developed a first-of-its-kind dataset from various Indian movie scenes. These scenes are classified as sexual harassment, sexual abuse, or sexual violence and exported in the PASCAL VOC 1.1 format. Our dataset is annotated on the identified relevant features and can be used to develop and train a deep-learning computer vision model to identify these issues. The dataset is publicly available for research and development.

Computer Vision and Pattern Recognition,Machine Learning

What problem does this paper attempt to address?

The aim of this paper is to develop a deep learning-based method for the early identification of sexual harassment behaviors in videos. Specifically, the researchers addressed the shortcomings of current technologies (such as Google Cloud Vision API and Clarifai API) in recognizing such content by constructing a specialized dataset to train computer vision models to automatically detect scenes of sexual harassment, sexual abuse, and sexual violence in videos.

To achieve this goal, the authors first extracted scenes containing these behaviors from 6 Indian films and had them annotated by social scientists. They focused on the facial expressions of the victims, the body language of the perpetrators, and other key factors such as unwelcome physical contact. Additionally, the authors analyzed the limitations of existing image recognition technologies in handling such content, pointing out that they often fail to effectively identify information related to sexual violence or sexual harassment.

The research contributions include:
1. **Creation of a new dataset**: Containing 500 images extracted from films, these images are meticulously annotated with information about the victims, perpetrators, and unwelcome physical contact.
2. **Evaluation of the effectiveness of existing technologies**: Experiments demonstrated that existing technologies like Google Cloud Vision API and Clarifai API have significant deficiencies in recognizing sexual violence and sexual harassment.
3. **Discussion of ethical and social impacts**: The authors also collected opinions from American university students regarding this technology, discussing its potential application value and ethical considerations.

In summary, the focus of this research is on improving the automatic recognition of sexual harassment behaviors in videos using deep learning technology, providing new tools and technical support for enhancing public safety and personal protection.