Description of original award (Fiscal Year 2016, $396,686)
With recent developments in communication and storage technologies, facilitating the proliferation of illicit media, law enforcement agents have to analyze overwhelming amounts of data in child pornography investigations. There have been attempts at automated child pornography detection using established computer vision and machine learning techniques, but these approaches had limited success in identifying children. In the past two years, the computer vision and machine learning research community has developed very effective 'deep learning' techniques that, if applied to the child pornography detection problem, would yield dramatically improvement in accuracy.
We plan to leverage our research on detecting child pornography performed in a previous NIJ project. This proposal addresses the problem of automatically detecting child pornography in videos, through the development of research in convolutional neural networks, a very powerful deep learning model. This project will result in the development of DeepPatrol, an innovative software tool to assist law enforcement agencies in investigating child pornography cases that will be designed to fit into typical law enforcement practices and workflows. DeepPatrol will be released free to law enforcement organizations. ca/ncf
- Gun Wars and Community Terrorization: Investigating Longitudinal Gang Violence in New Jersey from a Networked Perspective
- Examining Radicalizations Risk and Protective Factors: A Case-control Study of Violent Extremists, Non-criminal Extremists & Regular Homicide Offenders
- LC-QqQ-MS screening assay for abused drug exposure based on covalent peptide/protein modification