Resolution independent camera sens
Driving with autonomous vehicles adapts different cameras for different purposes. The application of the different types of cameras causes differences among sensor signals. The other situation that has to be considered is the sensor signal degradation. One of the symptoms of sensor signal degradation is the distorted or very limited size of resolution leads to a very tight region of interest. In this situation, it is a not surprising requirement is able to handle variable size of images input. We cannot have traditional learning models, because it does not handle changing the size of the input. We can resize or crop the pictures, but we also lose information. We developed a new Convolutional and LSTM (Long Short-Term Memory) based model which doesn't need fixed-size resolution. With the convolution, we can expand the information in between the dimensions and the LSTM layers can remember the previous stages.
