Did you know that the ML benchmark datasets we heavily rely on like MNIST, ImageNet, CIFAR, etc., can have thousands of errors in their labels? 😱 Check out labelerrors.com for a few examples.
Last week, I attended Snorkel AI's conference on #DataCentricAI - where one of my favorite talks was Cleanlab: AI for Correcting Errors in Any Dataset!
It was fascinating to know that what started as a graduate school project from an idea of finding an erroneous label in a benchmark dataset - of using #ConfidentLearning to find and fix label errors in ML datasets, is now being used to correct data at a large scale by companies like Microsoft, Tesla, and Google is fab 👏🏻
Also, cleanlab.ai is #OpenSource, no-code, and an automated solution!
Did you know that the ML benchmark datasets we heavily rely on like MNIST, ImageNet, CIFAR, etc., can have thousands of errors in their labels? 😱 Check out https://t.co/Ok0aa7JZWX for a few examples.https://t.co/KvOlEtbnGk
— Jigyasa Grover ✨ (@jigyasa_grover) August 5, 2022