In the article "Cyclist Intention Detection: A Probabilistic Approach", Stefan Zernetsch, Hannes Reichert, Viktor Kress, Konrad Doll and Bernhard  Sick propose a holistic approach for probabilistic cyclist intention detection.  A basic movement detection based on motion history images (MHI) and a residual convolutional neural network (ResNet) are used to estimate probabilities for the current cyclist motion state.