Person Identification: Created a CNN model using Pytorch and tensorflow by training over 10,000 images with different classes such as appearance, any type of handbag or backpack etc.Vehicle detection: Using YOLO v8 framework with more than 10,000 annotated images to train a custom model that helps in identifying cars from the existing surveillance infrastructure. Used deepSORT tracking algorithm in different region of interest (roi) depending on the camera angles.Headlight Detection: No lights detection by training a custom model using the YOLO v8 framework. Annotated more than 5000 images by hand. My model was a great improvement (f1 prediction accuracy of 90% compared to the existing systems f1 prediction accuracy of 50%)