Dataset contains microscope images of various plankton species from lake Baikal.
Medium article about our work and how dataset is obtained.
Images are obtained from microscope camera with same image background and backlight.
We provide polygonal labeling for object detection, segmentation and classification tasks.
Currently, we provide images 72 different species as well as category other
for image artefacts, background, etc.
File data.json
contains list of objects in images.
- Field
image
contains direct link for source image to download category
contains object classpoints
is a list of polygon coordinates; each point is represented bydict
withx
andy
coordinates of point, each coordinate is in span[0,1]
from left and top
{
"image": "https://storage.yandexcloud.net/baikal-samples/images/d30543d7-4881-4160-8c25-f014dddeffa4-New2/d07053cd-77ef-4410-8b4f-27607b24977f-21-09-14-12-57-27.jpg",
"category": "cyclotella minuta",
"points": [
{"x": 0.3097988874625588, "y": 0.9425188988731992},
{"x": 0.3662815575524176, "y": 0.9425188988731992},
{"x": 0.3662815575524176, "y": 0.9904435886464127},
{"x": 0.3097988874625588, "y": 0.9904435886464127}]}
Updated main dataset. Test dataset is now part of main dataset.
Test dataset added
First version of dataset