Table 3 Divisions of different datasets

From: Accelerating domain-aware electron microscopy analysis using deep learning models with synthetic data and image-wide confidence scoring

Data type

Set name

Image count (Feature count)

Experimental

All experimental under focused images

175 (13956)

Experimental Training

100 (8441)

Validation

50 (3998) *

Test

20 (1108)

Human Round Robin

5 (409)

Synthetic

Synthetic Training

37 (8826)

  1. “Data Type” indicates whether the images comprising a given set are experimental or synthetic images, and “Set Name” represents a dataset naming shorthand for convenient referencing throughout this work. For example, “Test” refers to the test dataset consisting of 20 experimental images which is used to evaluate both the Experimental and Synthetic models. *Validation sets were not ultimately used in any of the models report herein, see Section ‘Training of ML models’ for details.
Back to article page