PriyangshuGhosh/Landing_detection-ornithopter- — reverse-engineered prompt

Build me a small Python project for detecting branch like landing spots for an ornithopter from images. I want it to be educational, not just using a ready made deep learning model.

Please create a simple workflow where I can load a folder of images, convert them to grayscale, resize them, do basic thresholding or edge detection, and train a tiny neural network from scratch with NumPy to classify whether an image has a usable branch or rod like structure. Include forward pass, binary cross entropy loss, backpropagation, and gradient descent written clearly so I can learn from it.

I’d like a notebook for experimenting and a clean src version I can run from the command line. Add simple prediction visualization so I can see the original image, processed image, and the model’s yes or no result. Keep it beginner friendly with comments and a README explaining the math ideas at a high level and how to add my own training images.