A repository for work done on data preparation and model-building for commercial truck detection.

Below are high level description of the project:

• Target: detection and masking all commercial trucks from aerial images
• Data and labels: iSAID data with COCO labels, 'Large Vehicle' category only
• Model architecture: Mask RCNN
• Implementation: Detectron2

The exhibits are created on a cpu machine. To create a conda environment, update the conda_environment.yml file prefix as needed and create a new project environment as follows.

conda env create -f conda_environment.yml
source activate truck_detect
pip install -r requirements.txt

Detectron2 is Facebook AI Research's next generation software system that implements state-of-the-art object detection algorithms. It is a ground-up rewrite of the previous version, Detectron, and it originates from maskrcnn-benchmark.