EfficientNetV2-L
BLOOFINZ-2022
Eastern Indian Ocean
19 Cast Stations
PhytoVision
BLOOFINZ-2022
Goes Lab · Lamont-Doherty Earth Observatory · Columbia University
An AI-powered system that identifies and counts microscopic marine plants — phytoplankton — from ocean water samples. This dashboard shows results from the BLOOFINZ-2022 research expedition in the Eastern Indian Ocean, helping scientists understand ocean health and the marine food web that ultimately supports Southern Bluefin Tuna.
🌊
Library Images
0
↑ Library 1 + 2
🔬
Species Classes
0
8 in training
◎
Val Accuracy
0%
↑ Best ep19 / Stage 2
⚙
Parameters
118M
662K trainable
🚢
Discrete Casts
19
69,851 particles
F1
Macro F1
0.50
110× class imbalance
→
Explore the Dashboard
Click any section below to dive into detailed data and visualizations
19 Cast Stations
→
🚢
Cruise & Stations
Explore all 19 sampling stations across the Eastern Indian Ocean. View species composition at each cast, SST, salinity, fluorescence, the full geospatial cruise track, and depth profiles. Click any station card to open a detailed modal with 5 data tabs.
60.62% Val Acc
→
🤖
AI Model Performance
Deep-dive into the EfficientNetV2-L classifier. View training curves across 22 epochs, the confusion matrix, per-class precision/recall/F1 scores, and confidence analysis. Two-stage transfer learning from ImageNet weights.
12,315 Images
→
🔬
Ocean Biology & Data
Species dataset, ESD & morphological parameters, carbon biomass calculations (Menden-Deuer & Lessard 2000), optical & fluorescence properties, feature histograms from raw FlowCAM data, and the long chain finder for bloom dynamics research.
All Sections
→
📋
Full Dashboard
The complete single-page dashboard with all 14 sections visible in one scrollable view. Best for researchers who want the full picture at once. This is the original, fully-loaded version as first presented.
?
About This Research
Plain English overview of BLOOFINZ-2022 · phytoplankton · FlowCAM · deep learning
What Are Phytoplankton?
Phytoplankton are microscopic, plant-like organisms that drift in the ocean. Too small to see individually (most are 1–100 µm), they are the foundation of the ocean food web — producing about half of all oxygen on Earth through photosynthesis. Different species thrive under different conditions, so tracking which species are present tells scientists about ocean health, nutrient levels, and climate change.
What is BLOOFINZ-2022?
BLOOFINZ stands for Bluefin Larvae in Oligotrophic Ocean Foodwebs, Investigations of Nutrients to Zooplankton. In February–March 2022, researchers from Goes Lab (LDEO, Columbia University) and collaborators sailed the Eastern Indian Ocean aboard the R/V Roger Revelle. They collected water samples at 19 locations and imaged the phytoplankton with a FlowCAM instrument — capturing over 69,000 microscopy images.
📷
FlowCAM Instrument
A FlowCAM (Flow Cytometer And Microscope) automatically photographs particles in a water sample as they flow through a narrow channel. It measures size, shape, color channels, and fluorescence for each particle — generating thousands of images per sample that would take humans days to sort manually.
🧠
Deep Learning Classification
We trained an EfficientNetV2-L neural network on 12,315 hand-labeled FlowCAM images across 8 species. The model learns to distinguish Prochlorococcus, Nitzschia, Nanophytoplankton, and others purely from their visual appearance. It reaches 60.6% top-1 accuracy — with 98%+ top-3 accuracy on the validation set.
🌊
Why This Matters
Phytoplankton drive carbon cycling, oxygen production, and the base of the food web that supports fish, including Southern Bluefin Tuna. Automated identification allows researchers to rapidly analyze samples from expeditions, track changes over time, and understand how ocean conditions affect microbial communities — at a scale previously impossible manually.