Dataset
See information about the dataset collection process and labeling on the Stalk Detection page.
The dataset consists of 2667 images, totalling 7681 stalk instances. The dataset is split into 80%
training, 10% validation, and 10% testing data.
Each annotation file (for each data split) is in COCO format.
The dataset directory structure is as follows:
dataset
test
annotations.json
images
IMGS.jpg
train
annotations.json
images
IMGS.jpg
val
annotations.json
images
IMGS.jpg