ultralytics/yolov5 ? reverse-engineered prompt

Build me a Python tool for advanced computer vision using the YOLOv5 model. I need it to be fast, accurate, and capable of a few key tasks. The main one is object detection—finding and drawing boxes around objects in images or video streams. It should also be able to perform image segmentation to outline the exact shape of an object, and image classification to categorize the entire picture.

I want a complete project that includes scripts for training the model on a custom dataset, running detections on new images, and validating the results. A crucial feature is the ability to export the trained model into different formats like ONNX, CoreML, or TFLite, so I can deploy it on different platforms, including mobile devices. Please set up all the necessary Python requirements for this to work.