Raw Camera Data Object Detectors: An Optimisation for Automotive Video Processing and Transmission

Whilst Deep Neural Networks (DNNs) have been developing swiftly, most of the research has been focused on videos based on RGB frames. RGB data has been traditionally optimised for human vision and is a highly re-elaborated and interpolated version of the collected raw data. In fact, the sensor collects the light intensity value per pixel, but an RGB frame contains 3 values, for red, green, and blue colour channels. This conversion to RGB requires computational resource, time, power, and increases by a factor of three the amount of output data. This work investigates DNN based detection using (for training and evaluation) Bayer frames, generated from a benchmarking automotive dataset (i.e. KITTI dataset). A Deep Neural Network (DNN) is deployed in an unmodified form, and also modified to accept only single channel frames, such as Bayer frames. Eleven different re-trained versions of the DNN are produced, and cross-evaluated across different data formats. The results demonstrate that the selected DNN has the same accuracy when evaluating RGB or Bayer data, without significant degradation in the perception (the variation of the Average Precision is <1%). Moreover, the colour filter array position and the colour correction matrix do not seem to contribute significantly to the DNN performance. This work demonstrates that Bayer data can be used for object detection in automotive without significant perception performance loss, allowing for more efficient sensing-perception systems.