Deep Species Detection
A service to process citizen science image datasets and train deep learning models for species identification.

Copyright IFREMER
License
Other
AGPL-3.0
Main Characteristic
This module processes citizen science image datasets to prepare them for training deep learning models for species identification. It provides tools to clean datasets by standardizing bounding box formats and unifying redundant annotations to confirm the presence of individuals, enabling more accurate model training and analysis.
The training pipeline is designed for automation, allowing users to train a YOLOv8 model on their own data or use an existing model trained on Deep Sea Spy data, specifically for identifying Buccinidae and Bythograeidae taxa.
Technical Categories
Computer vision
Business Categories
Earth Observation
Last updated
14.01.2025 - 11:58