Distance education, as a strategic instructional method, offers students significant advantages in maintaining the continuity and coherence of their learning processes. This study aims to investigate the actions and behaviors of students in live sessions, and prediction of attendance patterns using artificial intelligence. The data of the study were obtained by quantitative techniques. Two live sessions were recorded, and a five-second clip of each participant was extracted from these recordings. The meticulous annotation of 17036 images was conducted for the purpose of classification. Some results of the study shows, the highest attendance rates were observed at both the onset and culmination of the live sessions. Furthermore, it was discerned that, on average, three out of every four participants consistently adhered to the course content, exhibiting a commendable level of dedication. Intriguingly, some students displayed an elevated level of engagement during the sessions by actively participating in discussions, responding to queries, or expressing enthusiasm. On the other hand, a portion of the participants demonstrated disengagement from the course material through various indicators, such as diverting their attention, lowering their gaze, or physically disengaging from the camera's field of view. To accurately distinguish between active and passive course attendance, cutting-edge deep learning-based transfer learning models were employed, leading to an impressive 92.0% success rate in classification. In conclusion, this study underscores the significance of distance education as an efficacious pedagogical approach in ensuring the uninterrupted progression of students' learning experiences. Through the application of advanced deep learning techniques, the research elucidates the diverse manifestations of student engagement during live sessions, providing valuable insights into their active involvement and disinterest.