The project

We ( @Raun @Parya ) participated in Google Landmark Recognition Challenge. We finished in top 100 on Kaggle.

We used ensample learning, CNNs and a whole bunch of preprocessing. Ranked in top 100; we classified more than 1.4 million images into 15K classes based on landmarks seen in them. In the pipeline, we used both classic image processing techniques and modern deep learning methods to tackle the complexity of object detection and landmark recognition.

The Code

Code can be found in this repository.