Metadata | Value |
---|---|
Classes | canopy,crown |
Machine Learning Task | object_detection |
Agricultural Task | crops_detection |
Location | USA |
Sensor Modality | RGB |
Platform | ground |
Input Data Format | JPG |
Annotation Format | coco_json |
Number of Images | 500 |
You can easily open the dataset in any application compatible with the COCO format. Here is an example of how to do that using FiftyOne.
- Download the dataset, and extract the archive.
wget "https://ghaipublic.s3.us-west-2.amazonaws.com/datasets/broccoli_annotated.zip"
unzip broccoli_annotated.zip
- You should end up with a folder structure similar to this one:
Project/
|--broccoli/
| |--coco.json
| |--image1.jpg
| |--...
- Create a virtual environment, and install the Python FiftyOne package.
python -m venv .venv
source .venv/bin/activate
python -m pip install fiftyone
- Create and run the following script.
import fiftyone as fo
import fiftyone.zoo as foz
dataset = fo.Dataset.from_dir(
dataset_type=fo.types.COCODetectionDataset,
data_path="./broccoli/",
labels_path="./broccoli/coco.json",
include_id=True,
)
# Verify that the class list for our dataset was imported
print(dataset.default_classes)
print(dataset)
session = fo.launch_app(dataset)
session.wait()