Guest
Artur Fattakhov, Ilya Kibardin and Dmitriy Abulkhanov share their winner’s solutions of Kaggle Camera Model Identification. In this competition, Kagglers challenged to build an algorithm that identifies which camera model captured an image by using traces intrinsically left in the image. From this video you will learn: - How to get additional photo data - Training Scheme with cyclic learning rate and pseudo labeling - Snapshot Ensembles aka Multi Checkpoint TTA - Training on small crops and finetune on big crops to speed up without loss in quality - Prediction equalization Slides: Github: Yandex hosts biweekly training sessions on machine learning. These meetings offer an opportunity for the participants of data analysis contests to meet, talk, and exchange experience. Each of these events is made up of a practical session and a report. The problems are taken from Kaggle and similar platforms. The reports are given by successful participants of recent contests, who share their strategies and talk about the techniques used by their competitors.
