A new method using deep learning and cameras can detect and predict when people are in a building. This helps control heating and cooling systems more efficiently, saving energy. By training a computer model with images of people in offices, the system can accurately identify activities like sitting, standing, walking, and napping. This information is used to create real-time profiles of how much heat people are giving off, which is more accurate than traditional methods. This can also help improve indoor air quality by adjusting ventilation systems based on occupancy. The study shows that this approach works well for detecting people in buildings and predicting their behavior.