Jacob Cordell

Data Science of Smart Cities through Vertically Integrated Projects

• Labeled 300+ images of road distresses using bounding boxes and image segmentation to be used for CNN training
• Created and optimized a UNET++ model on two datasets of images with the goal of maximizing F1 Score to predict crack segmentation in images
• Explored models with different frameworks such as UNet, DeepLabV3, FPN as well as different resnet encoders
• Continuing to work on a project involving determining accuracy of sensor data

• Company Overview: Corporate Partner: Purdue Student Life
• Utilized Agile framework
• Tested, validated, and reformulated models to foster accurate prediction of outcomes
• Created graphs and charts detailing data analysis results
• Cleaned and manipulated raw data
• Corporate Partner: Purdue Student Life