For this project, I made a computer vision model using a U-Net architecture  with PyTorch for vehicle detection. The images are from traffic cameras at several intersections in a major city. The entire dataset include images facing all four directions. The dataset is split into training and validation set based on the direction the traffic camera is facing.  I employed techniques such as gradient clipping, random drop-out and dataset augmentation when training the computer vision model. The resulting model achieved 0.7 IOU (Intersection Over Union) when comparing prediction to ground truth.