Search the Community

I've really enjoyed collecting and looking at forams under the microscope since I heard about them about 2 years ago. Identifying them can be very challenging for a novice like myself, so I thought I would look into getting the machines to help me. I've been a software engineer for the last 20 years so I felt comfortable diving in to some code. I tried writing my own image recognition model, using existing models (VGG16, VGG19, Res-Net50) but finally decided to use the code from a paper by Hsiang et al [2019] (https://agupubs.onlinelibrary.wiley.com/doi/full/10.1029/2019PA003612). The dataset was from http://endlessforams.org/ and contained planktic forams. I had to just crop them so only the foram was in the frame, not the border and text. Luckily a python script could do this for the ~30 000 images or it would have taken me ages!! After updating the python code to work with the latest Tensorflow, I trained it on the dataset and created a web interface for it with great results! I would find images online that it hadn't seen before and it got the correct species each time I tried. I would love to train it on some other species of forams, maybe some benthic (bottom dwelling) ones. I think it might also work on shark teeth, that would be really cool and I am sure there are many, many images that can be used to train it on. I made a video showing the process and how it identifies it in real time: I ran out of microfossil holders so 3D printed my own:

The first specimen is about 0.70mm long and the last one has a diameter of 1.4mm. I searched all 166 pages at marinespecies.org and came of two look-a-likes for the first specimen, they were, triloculina and quinqueloculina. The last specimen sorta looks like something named reophax subfusiformis. I can't say for sure if those are the correct species for these but it's a place to start. I've come across hundreds of these things and to date only about 3 can be identified by scientist.

These things are very small, my best guess is that they are well below 0.50mm. The bivalves and gastropods found in the sandstone nearby suggest that they are from the Pliocene period. There were a lot of these things found. I have researched these and found several photos but it not easy to tell which is which.