Grandmaster Series – How to Perform Large-Scale Image Classification

In episode two of the Grandmaster Series, learn how participating members of the Kaggle Grandmasters of NVIDIA (KGMON) built large-scale image classification models to win the Google Landmark Recognition 2020 Kaggle competition. 

In this landmark recognition challenge, the team had to build models that recognize the correct landmark (if any) in a dataset of complicated test images. This is easier said than done, given landmark recognition contains a much larger number of classes. For example, there were more than 81,000 classes in this competition. Watch the video to learn how more! 

Additional Resources: 

  1. Learn How How Kaggle Makes GPUs Accessible to 5 Million Data Scientists
  2. Competition Page: https://www.kaggle.com/c/landmark-recognition-2020
  3. A paper about this solution: https://arxiv.org/abs/2010.05350
  4. You can find the code here: https://github.com/haqishen/Google-Landmark-Recognition-2020-3rd-Place-Solution
  5. Learn more about NVIDIA Accelerated Data Science: https://www.nvidia.com/en-us/deep-learning-ai/solutions/data-science/ 

About the presenters: 

Host Jim Scott, head of developer relations, data science, at NVIDIA. Over his career, he has focused on enabling business to solve the most complex engineering and data related problems. His expertise in blending business needs with technology to drive innovation has influenced every major industry.

Gilberto Titericz, known as Giba is currently a senior data scientist at NVIDIA. Prior to NVIDIA, he worked at Ople, Airbnb, Petrobras, and Siemens. Gilberto had held the #1 position at Kaggle for more than two years. 

Dr. Christof Henkel, a Ph.D. in mathematics with a focus on probability theory and stochastic processes is a senior deep learning scientist at NVIDIA. He focuses on the fields of computer vision, natural language processing, graph neural networks, and reinforcement learning.

Bo Liu holds a Ph.D. in Applied Math and Statistics from Johns Hopkins University. Bo spent time working in Fintech and while doing so was competing in his free time and earning his Kaggle Grandmaster’s title. Bo joined NVIDIA in May of 2020 and is mostly interested in deep learning competitions, especially those related to computer vision.

Ahmet Erdem is currently a Senior Data Scientist at NVIDIA with a background in Computer Engineering and Artificial Intelligence. He has prior experience in robotics and software engineering. While he is active in a variety of competitions, his primary focus is deep learning on unstructured data.

    Leave a Reply