Discover advanced techniques for large-scale image classification in this 54-minute video from the Nvidia Grandmaster Series. Learn how Kaggle Grandmasters of NVIDIA (KGMON) built winning models for the Google Landmark Recognition 2020 competition, tackling the challenge of recognizing landmarks across 81,000+ classes. Explore classical approaches, winning solutions, validation strategies, code efficiency, and modeling techniques. Gain insights into third-place solutions, including architecture, fine-tuning, postprocessing, and ensemble methods. Delve into competition-specific strategies like cutout augmentation and label submitting. Led by industry experts, this comprehensive tutorial covers everything from introductory concepts to advanced topics like vaccine degradation, providing valuable knowledge for data scientists and AI enthusiasts interested in computer vision and large-scale image classification challenges.
Overview
Syllabus
Introduction
Welcome
Competition
Classical Approach
Winning Solution
Validation Strategy
Code Efficiency
Modeling
Third Place Solution
Architecture
Finetuning
Postprocessing
Ensemble
Competitions
Cutout Augmentation
Label Submitting
Vaccine Degradation
Taught by
NVIDIA Developer
Tags
Reviews
5.0 rating, based on 3 Class Central reviews
-
Arjun S PramodHi, this is my first time being watching an episode in the Grandmater series. I really enjoyed learning different approaches of dealing with such large-scale image classification tasks. This really interested me, I'll surely be joining for the further similar episodes.
Thank you -
Naveen Gokul SVery useful information and clear on topics thanks to person who made it clear
This course will be useful for me at future -
Thiago CamposPoderia ter uma demonstração como faz a classificação de larga escala, não só falar dos fundamentos como é como eles usam
