What is different between each Point generation method?

[palika w.]

Hi,

Firstly, thank you so much for your support. My question would be so basic, but I am so new about the CoralNet and analyzing the benthic cover data. I actually have two basic questions. 

1) What is different between Simple Random, Stratified Random and Uniform Grid? And I wonder which method is the most accuracy to analyze the data.

2) How many minimum annotation points should be in a quadrat image to make the analysis be more accuracy?

I really appreciate your answer.

Best,

Palika

[Stephen Chan]

Hi Palika, welcome!

Simple Random: Let's say our image is 4000 pixels wide and 3000 pixels high. For every point, we randomly pick a number between 0 and 3999, and we randomly pick a number between 0 and 2999, to determine the location of that point. If we are generating 20 simple random points, then we do this process 20 times.

Stratified Random: We think of the image as being divided into a grid of cells. For example, 4 rows and 3 columns of cells, for a total of 12 cells. Then, within each cell, we generate a certain number of random points. This guarantees that the points are somewhat evenly distributed throughout the image.

Uniform Grid: Again, we divide the image into a grid of cells. Then, within each cell, we place 1 point at the center of that cell. This arranges the points themselves into a grid pattern.

I'm not very well versed on statistical methods, so maybe someone else can better explain which kinds of studies benefit from which methods. But all three methods are fairly popular on CoralNet, so we have maintained all three options.

The number of annotation points per quadrat image also varies greatly between projects. In terms of analysis accuracy, I think the best number of points depends on how much area is covered in a single image, and depends on the subject matter you're analyzing. In terms of the annotation workflow, somewhere between 10 and 50 points is probably an ideal range to work with. Sometimes it can get trickier to distinguish different points in the annotation tool when you have 100 or more points.

[palika w.]

Hi Stephan,

Your answer is clear and it is really helpful for my research. 

Thank you so much. I really appreciate your help.

ในวันที่ วันอังคารที่ 5 ธันวาคม ค.ศ. 2023 เวลา 13 นาฬิกา 41 นาที 37 วินาที UTC+9 Stephen Chan เขียนว่า: