Didi-Udacity Self Driving Car Challenge 2017

Pedestrian detection

Cluster 3D Points using region growing clustering of depth map grid cells Filter out clusters with width and depth Select most probable cluster Track selected cluster

Car detection

Cluster 3D Points using region growing clustering of depth map grid cells Filter out clusters with width and depth Project clustered points on 360 degree panoramic view Predict boxes with pre-trained deep neural network Cluster boxes and select most probable box Track selected box