Developing a Chessboard Image to FEN Converter with Elixir

69
clicks
Developing a Chessboard Image to FEN Converter with Elixir
Barrett Helms presented his journey creating an Elixir-based web application designed for practicing chess tactics. His goal was to develop an image recognition system that could interpret photos of chessboard setups and convert them into Forsyth-Edwards Notation (FEN) for easy integration and interactive display within the app. While most resources for image recognition opted for Python, Helms was determined to use Elixir. He faced challenges dealing with poor accuracy and overfitting during image detection and classification, and needed to adjust his approach for a more general application. His talk covered aspects from grayscale image processing, edge detection using OpenCV, and leveraging Rust for parts of the task where high performance was crucial. Helms also explored the use of LiveView for real-time updates, NXTensor for data management, and Axon for machine learning. Despite initial setbacks and a presentation hiccup due to environment issues, Helms successfully demonstrated a working prototype with perfect input and stressed the importance of broader training datasets for real-world applications.

© HashMerge 2024