Challenge background
The cases of child abuse, domestic violence, theft, kidnapping, and terrorist attacks have significantly increased in Nigeria. Although efforts are being made to install surveillance systems in homes and communities, most of these systems record surveillance footage but are unable to detect abnormal behavior in real-time. Hence most criminal cases are only discovered after they have occurred. This project proposes deep learning as a solution to this challenge.
The problem
Surveillance systems in offices, streets, or homes should be able to detect abnormalities in the environment in real-time to prevent crimes from happening or stop ongoing criminal activities. This preventive approach to solving crime will greatly decrease the loss/ destruction of properties and prevent fatal injuries/ loss of life in extreme cases. Using AI, we can alert people of danger in time and thereby improve the security conditions in the country.
Goal of the project
1. Source for labeled surveillance video datasets for crime/ abnormalities
3. Prepare data and develop fast deep-learning anomaly detection models for real-time usage
4. develop an interface to deploy a successful model for possible integration with surveillance systems (with alert capabilities).
Project timeline
- 1
Week 1
Data Collection. The Team will focus on sourcing datasets for training and evaluating trained models.
- 2
Week 2
Data Cleaning. The Team will prepare the data collected for training.
- 3
Week 3
Data Analysis. The team will perform EDA and begin explore algorithms suitable for solving the given problem.
- 4
Week 4
Modeling. The Team will train models and evaluate performance.
- 5
Week 5
Deployment. The Team will consider alternative interfaces to deploy the trained model for use.
What you'll learn
Image & Video Processing, Anomaly Detection, Object Detection.