EZOBDE | EaZy Object Detection

This tool was developed to easily train an object detection model (Faster R-CNN) from ImageBox labels without the need to pre-download images or label JSON.

Edit YAML file

The YAML file requires the following fields, which can be retrieved from Labelbox:

1. api_key - An API key from Labelbox
2. project_id - The project ID from Labelbox
3. classes - The class name used in Labelbox, e.g. plant, panicle

Run code

Python

To run the code, use:

./detecto_labelbox_object_detection.py -y config.yaml

Singularity

To build the container, run:

singularity build ezobde.simg docker://phytooracle/ezobde:latest

Then, run the container using:

singularity run --nv -B $(pwd):/mnt --pwd /mnt ezobde.simg -y <your YAML file here>