Classify

---

I created this visual classification and annotation tool to support AIMOOC, a project exploring machine learning for organ-on-a-chip cultivation. I wanted to understand not just how biologists classified microscopy images, but which features informed their decisions. The aim was to make that expertise explicit and help organise the information needed for the machine learning work.

The application lets researchers create image-based studies, define classes, and collect classifications, comments, and visual markings from participants. The Java version uses a Spring Boot REST API, relational storage, and a browser interface. It brings responses together through accuracy breakdowns, comments, and heatmaps, making individual judgments easier to compare.

The main challenge was turning subjective visual interpretations into structured data without losing the reasoning behind them. Image-relative marking grids kept annotations consistent across display sizes, while a straightforward workflow helped keep attention on the images. The project reinforced that useful machine learning starts with understanding the data—and that an expert’s explanation can be as valuable as the label itself.