The biggest challenge that we ran into was how to detect the blinks and how to count them. The code had some inherent bugs which could not detect blink correctly. The code was built on the idea that once the eye is closed and opened, it is registered as a blink. But we had also planned on detecting if the person closes their eyes for a long time to detect drowsiness or sleep. So, to correct this, we set a threshold for the detection of the number of frames to reset the blink counter condition and still could register blinks. The EAR formula which was refered in a research paper didn't have an easy implementation. We had to search for ways to properly detect the points in the eye so that the formula could be applied. After extensive research, we had discovered a library called dlib which helped us detect features from our face, given some points which generally identify a face. So, combining this library with our idea, we came up with a solution. Another problem we came up with was to set the treshold for detecting that the eyes were closed. This had to be corrected with multiple trial and error methods and setting threshold till we got one which worked perfectly.