Implementation of Automated Small Object Detection System with Deep Learning. The objectives of this project are
- to evaluate several deep learning architectures over small object-based datasets
- to integrate the detection pipeline with Cytomine, an open-source web platform which fosters collaborative analysis of very large images and allows semi-automatic processing of large image collections via machine learning algorithms.
The works is currently heavily based on Detectron2, a deep learning-based detection framework developed by Facebook, and powered by Pytorch.
This project is released under the Apache 2.0 license.