Machine Learning for the Fast and Accurate Assessment of Fitness in Coral Early Life History

A machine learning pipeline was developed to measure three key indicators of coral juvenile fitness rapidly and accurately: survival, size, and colour.Two substrates were used to classify pixels through a machine learning image analysis pipeline to quickly identify and measure coral juveniles:a) field deployed terracotta tiles -1200 images with three different time point in the field -deployed at Davies reef in the Great Barrier Reef for a year-long experiment (see Quigley et.al)b) laboratory-maintained PVC plastic slides. -900 slides with images taken at 12 time points in the laboratoryBoth substrates used coral juveniles from the species Acropora tenuis within the first year of life. For both sets of images, measurements were taken manually for proof of study. Size was calibrated using the scale bar present in each image. Colour of juveniles was assessed using the CoralWatch Health Chart and was matched to the closest score on the “D” scale by a single person to minimise observer bias. Survival of juveniles was classified by eye as either alive or dead.Machine learning image analysis pipeline was developed for measuring coral survival, size, and colour. Assessment of manual vs. pipeline calculations of coral juvenile colour, size and survival, as well as time comparison between manual vrs pipeline measurements, were carried out. Further details are presented in the publication Macadam et al. (2021).

本研究开发了一套机器学习流水线，可快速且精准地测量珊瑚幼体适合度的三项核心指标：存活率、体型大小与体色。
本研究采用两种基底，通过机器学习图像分析流水线完成像素分类，以快速识别并测量珊瑚幼体：
a) 野外部署的赤陶瓦片（terracotta tiles）：
- 采集自大堡礁戴维斯礁（Davies Reef）为期一年的野外实验，包含3个不同时间点的1200张图像
- 该实验于大堡礁戴维斯礁开展，实验详情参见Quigley等的研究
b) 实验室养护的聚氯乙烯（PVC）塑料载片：
- 包含实验室环境下12个时间点拍摄的900张载片图像

两种基底均使用了孵化一年内的细枝鹿角珊瑚（Acropora tenuis）幼体。两类图像数据集均通过人工测量以验证研究可靠性。体型大小通过每张图像自带的比例尺进行校准。幼体体色采用珊瑚观察健康图表（CoralWatch Health Chart）进行评估，由单一研究者匹配至“D”标度下最接近的分值，以尽可能降低观察者偏倚。幼体存活率通过肉眼判定为存活或死亡。

本研究开发了机器学习图像分析流水线，用于测量珊瑚幼体的存活率、体型大小与体色。同时开展了人工测量与流水线计算结果在珊瑚幼体体色、体型及存活率上的对比，以及人工与流水线测量的耗时对比。详细研究内容已发表于Macadam等（2021）的论文中。