Have you ever struggled with identifying cooling holes in a dataset, only to find that they deform and split into smaller holes before fully blocking? I know I have. Recently, I tried using YOLO to identify these holes, and while it worked reasonably well, I realized that I needed a way to ensure that hole number 1 is indeed hole number 1 in any dataset.
The problem is that these holes can change shape and size, making it difficult to accurately identify them. I’ve tried using K-means clustering, but I’m only getting around 20% accuracy. Has anyone else faced this challenge? What methods would you recommend to improve hole numbering accuracy?
The data I’m working with is a series of xy matrices, with each hole being a single matrix. I’m open to any suggestions or ideas you might have.