Lyft Motion Prediction Competition

Our first competition team has competed in the ‘Lyft motion prediction competition’. It is a truly exciting competition since it acts in the revolutionary field of autonomous cars. This exact Kaggle competition is about predicting the behaviour of all the road users based on a map of the surroundings of a car including all the moving objects on the road.

Lyft is a ride sharing company similar to Uber. They issued the Lyft Motion Prediction for Autonomous Vehicles challenge. Lyft is planning on introducing autonomous drivers for their vehicles. To accomplish this, they collected a huge amount of data from traffic situations their drivers faced and organized a competition with it on Kaggle (a website with many large datasets and competitions). The goal of this competition was to accurately predict the motion of other traffic agents, such as other cars or pedestrians.

Most of us started with little practical knowledge of deep learning models and everything else that comes into play when using AI. During this competition, we learned to use PyTorch and set up a ResNet34 model (which consists of 34 hidden layers), bringing us to the top 10% of the leaderboard! Come talk to us in the Software Engineering channel if you want to hear more about this!