Procedural Modelling for Synthetic Data Generation in Automotive Applications

In order to monitor the vehicle transportation and notify the individual about the information of the obstacles based on the analysis performed in the MPEG4 video file. The parameters to predict the severity of the impact has also been analyzed. An approach to generate synthetic annotated information about the obstacle has been formulated. The information obtained could then be used to train deep neural networks in computer vision tasks. This procedural modeling approach enables physically accurate image synthesis obtained from hand modeled virtual worlds. A sample MPEG4 Video file of vehicle transportation is provided as initial input to the workspace for experimental purpose. Image analysis has been done and based on specific characteristic feature; the actual object present is identified. The same approach would be used in automotive applications in near future.