On Performance Measures During Behavior Detection Algorithms Implementation - Case Study

During the last few decades, the development of information and communication technology, Internet of Things (IoT), Big Data and AI has brought several unique tools with extraordinary added value. One can easily see that all these technologies and tools are now available via standard services at home, in cars, in a public life and during the social interconnection with the environment. Machine learning is a technology that computers can learn on their own to create and predict models. Furthermore, deep learning is a field of machine learning using deep neural network theory, using the principle of neural network corresponding to the human brain. Expansion of cities and population growth necessitate utilizing automated video surveillance and computer vision-based analyses for pedestrian and vehicle safety together with the growth of number of cars and traffic infrastructure [1]. All these aspects of modern technology give us several questions. How can we be so sure that all these “smart” algorithms work reliably? In this work, we introduce a case study for predicting an object behaviour with respect to its safety. Predicting trajectory of an object, either being detected or occluded, provides to predict all probable risky situations by exploiting the last seen parameters of the object movement even when it disappears. All the scenes, including potential collision situation, rely on object detection and tracking. One question remains! How can we measure the performance?