Trainable, multi-purpose detection module (Angryk, Banda, Martens, Schuh): This module can be trained by example to detect any type of solar event or phenomenon. The figure above shows a full-disk Hα image segmented into grid-based cells, with training examples on the left and detected events on the right. The cells in the left image are labeled: filament (yellow), no-filament (green), and discard (blue). The cells in the right image correspond to: true positive (green), false positive (blue), and false negative (red) detections. We find an initial overlap of over 80% compared to detections by the AAFDCC module, without any fine-tuning of the trainable module.