Text this: Fine-tuned RetinaNet models for Vision-based Human Presence Detection