Word Search Solver:

A Python script that automatically detects and solves a Word Search puzzle from an image.

Features:

Image Preprocessing: Automatically processes an input image to isolate the puzzle and letters.

Letter Bound Detection: Finds the letters' borders using a histogram-based algorithm.

Letter Recognition: Uses a custom-trained CNN to identify the letter in each bound.

Puzzle Solver: Implements a brute-force algorithm to solve the detected puzzle. It uses a word list which removes the necessity for the user to input most words.

Current Limitations:

Straight Images Only: The current version of the grid detection works best on images that are straight and un-rotated. Support for perspective warping is a potential future improvement.

Word Length: The word list consists of words at least 5 letters long. It does not detect any words that are shorter than the limit.

Maintenance Status: This is a personal project and is not actively maintained. Updates and bug fixes may be infrequent.

Requirements:

Pillow (If you plan on creating your own dataset)

PyTorch (The TensorFlow version works but is not as good as the PyTorch version)

OpenCV

NumPy

License:

This project is licensed under the MIT License. See the LICENSE file for more details.