Thanks! Right now the plan is to cover segmentation networks and inference optimizations (so that the trained network runs faster). We will probably leave object detection with YOLO or SSD as an optional exercise, but that might change.