Second Annual Data Science Bowl solution

Due to organizational mistakes in second phase our team highly overfitted and finished on 168th place.

Without making such mistakes we had chances to place near to the 50th position.

Here is plot of our predictions vs. groundtruth on validation subset.

Main contributions

We developed 2d landmark detection for SAX and 2CH left ventricle projections. Landmarks drawn with red.

Here is an animation of landmark tracking on SAX view.

We also made some experiments on 3D reconstruction using MeshLab software.