Forecasting Future Instance Segmentation with Learned Optical Flow and Warping
This work addresses safety in autonomous driving by enabling scene dynamics prediction, but it is incremental as it applies known optical flow and warping techniques to a specific task.
The paper tackles the problem of predicting future instance segmentations for autonomous vehicles by using autoregressively forecasted optical flow fields to warp current segmentations, achieving effective results on the Cityscapes dataset.
For an autonomous vehicle it is essential to observe the ongoing dynamics of a scene and consequently predict imminent future scenarios to ensure safety to itself and others. This can be done using different sensors and modalities. In this paper we investigate the usage of optical flow for predicting future semantic segmentations. To do so we propose a model that forecasts flow fields autoregressively. Such predictions are then used to guide the inference of a learned warping function that moves instance segmentations on to future frames. Results on the Cityscapes dataset demonstrate the effectiveness of optical-flow methods.