High Performance Camera-based Modules for Autonomous Vehicles

Vision based modules are essential to realize artificial intelligence robotics, smart traffic and driverless cars. Our research work focus on traffic sign recognition and object detection in context of vehicle and pedestrian detection. Several detection techniques are based on deformable part models (DPMs) and convolutional neural network (CNN) for its foreseen precision. We proposed multiple approaches to improve the performance of DPM and the more accurate DeepPyramid DPM (a mapped model into convolutional layers). Multiple techniques are employed to build our frameworks including but not limited to Fast Fourier transform, feature approximation, edge boxes, neighborhood aware cascade and region proposal networks Our fastest technique results show that we process Pascal VOC 2010 at 47HZ and VGA image at 40HZ with loss of 2% in precision.