Predicting behaviour with computer vision

With the development of artificial intelligence/deep learning approaches and the acceleration of computational power /hardware upgrades, a significant increase on researches is observed in the field of computer vision. In the light of these developments and the state of art techinuques, it has become possible to get important information from videos. This information, recognition of human actions and behaviours is undoubtedly valuable . The classification of human actions has found a wide range of uses from health to sports, from entertainment to robotic studies, from education to retail sector, from transportation to cinema. In this study, the classification of human movements and behaviors in the office environment, which cannot find much use in the literature, is discussed. Within the scope of the study, a new dataset has been created and the classification of human actions and behaviors has been made on the dataset. Successful classification algorithms and deep learning approaches has been preferred while making classifications. In addition, by extracting new features derived manually from the existing data, the success of their classification has been tried to be increased and the experimental findings have been discussed. The experimental findings obtained have been observed at a satisfactory level with results up to 0.95 accuracy.