cmusatyalab/OpenTPOD ? reverse-engineered prompt

Build me a web app for creating custom object detectors without writing code.

I want a simple interface where I can upload and manage training images, label objects using an integrated annotation tool like CVAT, then start training a model with one click. It should support common object detection models such as SSD MobileNet and Faster RCNN, and let me fine tune them on my own data. After training, I want to export the model so it can be used for inference with TensorFlow Serving.

Please make it feel like a full end to end workflow for non experts, from dataset setup to annotation to training to export. A clean dashboard for projects, datasets, jobs, and results would be great. If you need to look up current docs online, feel free to do that.