Investigating classification learning curves for automatically generated and labelled plant images

TL;DR

This study presents a plant image dataset and analyzes classification learning curves using ResNet, revealing power-law relationships and effects of label noise and model complexity on training data requirements.

Contribution

It introduces a new plant image dataset and investigates how label noise and model size influence learning curves in plant classification tasks.

Findings

Learning curves follow power-law relationships.

Label noise increases data requirements for the same performance.

Reducing trainable parameters also raises data needs.

Abstract

In the context of supervised machine learning a learning curve describes how a model's performance on unseen data relates to the amount of samples used to train the model. In this paper we present a dataset of plant images with representatives of crops and weeds common to the Manitoba prairies at different growth stages. We determine the learning curve for a classification task on this data with the ResNet architecture. Our results are in accordance with previous studies and add to the evidence that learning curves are governed by power-law relationships over large scales, applications, and models. We further investigate how label noise and the reduction of trainable parameters impacts the learning curve on this dataset. Both effects lead to the model requiring disproportionally larger training sets to achieve the same classification performance as observed without these effects.