Traffic Map Movie Forecasting 2021
- Study high-resolution multi-channel traffic movies of entire cities with road maps.
- Predict future traffic flow across spatial and temporal domain shifts.
- Discover underlying rules and patterns in movies or on graphs.
Building on its success at NeurIPS 2019 and NeurIPS 2020, Traffic4cast is going into its third year offering new challenges and opportunities. This year, we are challenging world-class machine learners to build traffic models robust to domain shifts in space and time. Our Traffic Map Movies have been derived from more than 1012 data points and cover 10 diverse cities around the world from 2019 through 2020, providing an order of magnitude more data compared to last year. The Traffic Map Movies contain eight dynamic channels encoding traffic speed and volume per direction and nine static channels encoding the properties of the road maps. In the core challenge, the models need to adapt to a drastic temporal domain shift due to Covid-19 pandemic. In the extended challenge, the models must predict traffic flow in entirely new cities based only on few test examples.
Interested in understanding traffic around the globe in a new way? Join us to help shape this year’s competition.
Highlights from 2019 and 2020
- Our chosen representation was highly effective and — as independent work has shown for precipitation prediction — should be considered a promising new technique for tackling complex geo-spatial processes.
- Our 2020 competition asked participants to predict traffic in the second half of 2019 from given data in the first half of 2019. Thus temporal transfer learning across seasons was possible, although a performance boost could be achieved when also using the validation set data provided for training.
- The work of H. Martin et al. following their 2019 prize winning solution indicated that some degree of transfer of learned traffic patterns to almost unseen cities was possible with their GNN approach.