Implement Machine Learning Model for Threat Detection in X-ray Images
Overview:
Our customer, who specializes in security baggage scanners and physical security, has a set of baggage scanner machines deployed at strategically important locations. They want an improved machine learning model that can detect threat objects like knives, scissors, and guns in x-ray images.
Challenges:
- Understanding the setup of the framework (Alexey Darknet) needed to train the YOLO model.
- Understanding the methods needed to create a perfect dataset for training such a model.
- Validation and testing of models.
Development of a C# API for threat detection using the trained model.
Solutions:
- We developed the YOLO model and its corresponding API, which helps to detect threats using trained models.
Our solution consists of:
- A trained model with good accuracy for threat detection in x-ray images.
- A C# API that aids in threat detection.
- A demo test application to evaluate the model on images.
Outcome:
Our solution helped our client introduce a new threat detection feature with the help of a trained YOLO threat detection model.
Technical knowledge and understanding:
- Understanding of the YOLO model and Alexey Darknet.
- Understanding of CUDA version and its dependencies when building Alexey Darknet.
- Understanding of model validation and OpenCV library.