GitLab

Hi there, I was wondering if it's possible to add a new feature for showing streaming AIS data on the web interface of AISDB. The idea would be to use this feature to integrate with vessel tracking machine learning models. This way, we could see the true vessel trajectory and compare it with the predicted one.

Ideally, I believe we need asynchronous communication between the model and the interface. The model will probably output faster than what's reasonable for the interface to handle. We need a pool to store the model's output in this case. The output is sets of NumPy arrays where each row is a vessel, and each column is a pair of coordinates (lat and long). Another possibility of output would include the depth with the coordinates. The positioning of the vessel will be sorted by time.

Another feature would be showing the trajectory improvement over the network's training. As the model improves over the training, the outputs will change for the forecasting and not for the true data. This means we will plot the same data repeatedly, but hopefully, the forecasting will be closer to the true data behavior after every iteration.

In Fig 8 and 9 of this Article, there's an example of contrasting trajectories. Our case, however, would be shown in a Mercator projection.

Let me know if that's possible to be implemented. Thanks for the help!