Machine Learning & Citizen Science & Conservation Research
Flukebook applies computer vision algorithms and deep learning to identify and track individual whales and dolphins across hundreds of thousands of photos. We help researchers collaborate with each other and citizen scientists contribute to the effort. A.I. scales and speeds research and conservation.
Step 1. Deep Learning Finds Animals
We train computer vision to find individual whales and dolphins in photos and identify the species.
Step 2. Algorithms and Neural Networks Identify Individuals
When we know where each animal is, we can identify them individually using algorithms that make digital "fingerprints" for each animal, such as identifying them by their unique body coloration or fin edges. We replace hours of human labor with just a few minutes of computer vision, scanning for matches across tens of thousands of photos.
Step 3. Population Dynamics Define Conservation Action
If we can quickly track individuals in a population, we can model size and migration to generate new insights and support rapid, data-driven conservation action.
One Platform, Many Species, Many Researchers
We can identify individuals of these species using fully automated computer vision:
Humpback whales (Megaptera novaeangliae)
Sperm whales (Physeter macrocephalus)
Orcas (Orcinus orca)
Bottlenose dolphins (Tursiops truncatus)
Spotted dolphins (Stenella frontalis)
Right whales (Eubalaena sp.)
Fin whales (Balaenoptera physalus)
Humpback dolphins (Sousa plumbea)
...and more soon...
Megaptera novaengliae (humpback whales) can be identified by the contrasting color and the trailing edge of their flukes using three algorithms: PIE, Kaggle 7, HotSpotter, CurvRank, and OC/WDTW. Learn more.
Physeter macrocephalus (sperm whales) can be identified by the trailing edge of their flukes using two algorithms: CurvRank and OC/WDTW. Learn more.
Tursiops truncatus(bottlenose dolphins) can be identified by the trailing edge of their fins with the CurvRank algorithm. Learn more.
Eubalaena spp.(right whales) can be identified by the their heads, peduncles, flukes, and more using the PIE, HotSpotter and DeepSense.ai algorithms. Learn more.
"Sperm whales roam so vastly that no one research group can study them across their range. PhotoID as a tool for conservation and research finds power in numbers and international, inter-institutional collaboration. Flukebook enables us to do this easily." - Shane Gero, The Dominica Sperm Whale Project