Overview / Usage

Zooplankton directly influence the filtering of waters and ecosystem balance. They are great sources of food for fish and crustaceans, sources of food for us.
Identifying and measuring the growth of zooplankton are laborious, manual tasks done only by specialists. There are currently expensive hardware and software that help with this task, but they do not work well for all scenarios.
This project intends to automate these tasks and help accelerate important research that is not developed because of its difficulties, using deep learning to transfer knowledge from a specialist to a computational model.
Conventional methods used by area researchers are used to determine the size of zooplankton after being recognized by the model.

Methodology / Approach

The data set are microscopic images of the specialist himself, currently being labeled by him. Some algorithms of object recognition, such as sliding-window, R-CNN, YOLO, Faster R-CNN and SSD, will be tested, identifying the algorithm with the greatest accuracy for this problem.

Technologies Used

Intel Distribution for Python
TensorFlow* Optimizations for the Intel Architecture