Computer Vision AI Project (Roboflow)
Problem and Goal:
Monitoring lab personnel for PPE compliance (specifically safety goggles) is often manual, inconsistent, and prone to error. The goal was to build a computer vision model that can automatically detect goggle use in laboratory images.
Tools & Methods:
Roboflow for dataset creation and annotation
RF-DETR (Nano) model architecture for object detection
70/20/10 dataset split for training, validation, and testing
Iterative training with confidence threshold adjustments
Results:
Early-stage model detects goggles with baseline accuracy on test images
Sample outputs show bounding boxes identifying goggles in lab settings
Workflow includes annotated dataset, training screenshots, and detection examples
Business Impact:
This project demonstrates hands-on experience in computer vision workflows applied to lab safety compliance. Once optimized, the model could automate PPE monitoring, support compliance audits, and reduce manual supervision — increasing both efficiency and accuracy in lab safety environments.
Real-world example from my time at Bayer in a cGMP lab where safety goggles were required PPE. One image shows the model detecting and labeling goggles, while the other confirms no goggles detected — validating the workflow at a 95% confidence range:
