Large Scale Datasets for AI: How to achieve quantity and quality ( Talk on Analytica 2022)
I gave this talk on Analytica 2022. Since the talk itself was not taped, I decided to tape the rehearsal and just put it online. This is the outline: 00:28 Overview 01:55 The effect of largescale datasets 05:04 Largescale datasets in histopathology 09:48 How much data is enough 11:36 The proper annotation tools 12:50 Can t some students annotate the data 14:45 Simulation: How many annotators do we need 18:27 How does the number of annotators affect the deep learning model 19:54 Can an algorithm help in the annotation 25:00 A proposed framework for annotation 27:30 Summary: How to create largescale datasets 28:08 How to validate on largescale datasets 30:42 A study register for machine learning 32:59 Summary: How to validate 33:48 Summary of the talk
|
|