Hey everyone! I just published a writeup on a project where I applied what I learned in Course 4 of the Deep Learning Specialization to a real problem: extracting chess positions from screenshots.
The biggest takeaway: the course material mapped 1:1 to this project. Conv layers, pooling, batch norm, data augmentation, overfitting diagnosis — all of it showed up. Building something real alongside the coursework made the concepts stick way harder than assignments alone.
Blog post (full journey including failures): Screenshot → FEN: Teaching a CNN to Read Chess Boards — Andre Batista
Would love to hear if anyone else has applied Course 4 to hobby projects!