Collecting more annotated data

Hi all!¶

I hereby invite you to help annotate more DIOPSIS images to expand the available labelled dataset.

Why collect more data?¶

As you might know, the best way to improve the accuracy of a machine learning model is by adding more data. As more data provides a model with more examples for accurate detections and classification predictions, which will help your DIOPSIS classification and detection model. What’s more, for my Data Science Master Thesis I am investigating the effect of including annotations from annotators with varying expertise on a model’s insect classification performance. Hence, the collected data would massively help my thesis.

What is in it for you?¶

The annotated data will be made publicly available, such that everyone in the challenge will have equal access to it. So, you can use it as extra data to train your detection and classification models for the DIOPSIS challenge. We will share the extracted data after a couple of weeks of data collection.

What is expected of you?¶

If you choose to help with annotation, first of all: Thank you! For the annotation process we will ask you to annotate full DIOPSIS images (for example see below). For each of these images we ask you to draw bounding boxes around each insect in the image (this will serve as new data for the detection algorithm) and then label each insect according to the DIOPSIS ontology (this will serve as new data for the classification algorithm). The more images we annotate, the bigger the new dataset will be!

Example DIOPSIS image with annotations

When can I expect what?¶

We plan to follow the following timeline for the publication of the annotation tool:

Week 19 → Make annotation environment available

Week 19 - 23 → Run data collection

Week 23/24 → Make collected data available

Any questions?¶

Feel free to contact me if you have any questions about the data collection or annotation. My email: heleen.visserman@ru.nl.

Drop a message below this topic if you are interested in helping the annotation project and follow the topic to keep posted on updates when the annotation project is live.

Kind regards and happy annotating,

Heleen Visserman