Human errors are undoubtedly the pain points in railway operation, raising concerns about safety and services. As one of the world’s leading railway operators, MTR is always exploring innovative opportunities that help improve the safety and services. Using system data to monitor operators’ action can prevent these incidents, however interfacing with mega systems like signalling and power control systems requires expensive modification. The latest video capturing technology enables cost-effective collection of data directly from the screens of operators without affecting the system operations. It also facilitates the design of a generic interface for those mega systems. Additionally, video analytic techniques like deep learning and text recognition are performed on the video streams to alert wrong actions from operators. We collectively name this type of application as Smart Watchdog for Human Factors, with two examples as follows. 1.Railway Route Setting Monitoring System interfaces with signalling system to provide instant warning when an intention to violate the predesigned route setting rules is detected by deep learning. 2.Power Switching Communication Enhancement System interfaces with power control system to enhance the communication during power switching operations and alert operators when wrong power switching execution is detected by text recognition.