kaizen49/object-detection-yolov8 ? reverse-engineered prompt

Build me a small object detection demo that compares YOLOv8 and Faster R CNN side by side.

I want a simple local web app where I can upload an image and see both models detect objects on the same picture, with bounding boxes, class names, confidence scores, and a quick comparison of how long each model took and how many objects it found. It should also support live webcam detection in the browser, with a choice to switch between the two models and a confidence slider.

Please also include a notebook version for deeper testing on sample COCO images, with side by side outputs, basic charts, and an exported results file. It should use pre trained weights and download anything needed the first time it runs, and use GPU automatically if available but still work on CPU.

The app.py file expects a frontend template, so please create that too, along with any needed static folders and a clean, simple interface. If you need to check current library docs online, go ahead.