I am interesting

Dear Maria

i never perform fish egg counting but in principle

you can try to use imagej software

https://imagej.nih.gov/ij/

in the same way that is used for perform cell counting

https://youtu.be/8c68qIz_ftw

good luck

Manuele

Thank you so much, Manuele! I will try this program, and hopefully it will work, thereby saving me a lot of time.

An old fashioned method is as follows: Make a grid template which is about 1/20th the area to be counted from. Place it over a few (3 or 4) random spots and physically count them to get an average count per grid. Then see how many grids will fit into the actual image (it will be close to 20) and do the simple multiplication.

Similar methods have been used in estimating bird counts in sanctuaries.

Thank you for your suggestion, Srini Vasan! I could calculate an average count, but I need accurate counts to calculate the hatch rate, and the incubators where I kept the eggs are not that large (diameter: ~20 cm). Say that I have around 500-600 eggs in each incubater in the beginning of hatching. In addition, the eggs are in one layer, but they are by far evenly distributed. I will test the method with a few pictures to see how close to the truth I get.

I have also tested using ImageJ to analyse my pictures, but I'm unable to create good binary pictures due to the grid in the bottom of the incubator. The material is almost the same tone as the eggs (however different in color). When I change it to grayscale their tones overlap. It doesn't matter where I set the treshold, there is too much noise anyway.

If there was a way of telling the program that "these range of colors, around this size and almost perfectly circular" are the units I am looking for, then I think it would work. This shouldn't be impossible to do, seen in the light of how far we have come with machine learning. However, I don't think recognizing fish eggs are on top of the computer scientists priority list. I am aware that there exists programs where you can teach the program to recognize different cell phenotypes, but I don't know how it would work with a noisy background that is not black or white. I don't want to purchase any programs that I don't know will work...

Dear Marie

looking in literature for Free software alternative ti imagej i just found this paper

https://parasitesandvectors.biomedcentral.com/articles/10.1186/s13071-016-1870-1

where the authors use their software named icount for counting egg from mosquitos.

i suspect that the size is different from your and it may surfer of limitations similar to imagej but the tool could be downloaded for free at https://www.dropbox.com/sh/28vm6tv7wrm76ln/AADsCz8lQB37qU-n8j2qGjw_a?dl=0

so if you have time you can test also it.

good luck

Manuele