Effective Semantic Video Classification Model for Driverless Car

The prime motto of the research work is to design a semantic video classification model with improved efficiency. In order to attain this preferred target, the problem statement is concisely framed to detect the situation in which the brake of the preceding car is applied or not. The semantic video classification model is proposed using transfer learning techniques for autonomous vehicles. To implement the proposed model, a database consisting of positive and negative images is used to train the machine learning architecture. For effective implementation of the model, the car as an object is detected in the frame of the video using the HAAR cascade approach, and subsequently, the brake light status is scanned for estimating the possibility of brake applied, brake not applied, left indicator activated, right indicator activated, parking activated and light off state. Further, the actual time required to classify the video semantics is calculated. The event detection time of the proposed model is less than 0.69 ms with 99. 23% of accuracy, and precision along with 100% F1-Score and sensitivity. Further, a cogent comparative analysis of the model was carried out by using the performance evaluation with respect to the modelling style.