This study aims to assess the visual engagement of the video lectures. This analysis can be useful for the presenter and student to find out the overall visual attention of the videos. For this purpose, a new algorithm and data collection module are developed. Videos can be transformed into a dataset with the help of data collection module. The dataset is prepared by extracting the image frames from the video and marking them with a number of faces, the number of eyes, the status of eyes and the engagement score along with nominal values of engagement level. This data is transformed into time-based data items by using the attribute number of frames processed per second (PFPS). A case study for the assessment of TEDx video (length 8 minutes and 53 seconds) is included to validate the results and to extract statistical information from the dataset. Frames in the video are 16047 and they are transformed into 2675 keyframes. Machine learning classifiers are applied for the analysis of the dataset. The findings of this analysis help the presenter and the student to measure the quality of the visual content of the videos without actually going through it.