DEXTR

askdasjkdjakshdkjhaskjdhaskhkjashkjhasd

DEXTR is an AI-powered tool that is available to you from the start. It works by you placing four markers (or more) - one on the most northern edge of what you want to label, and one on the most eastern, northern, and western edges too. It is a powerful tool that can get you really good results in just four clicks.

You need to give DEXTR at least four markers for it to work. You create these markers by left-clicking or by pressing "A" while your mouse-pointer is hovering over one of these edges. After you have giving four markers, the DEXTR-model will "look" for any object inside of those four markers.

If you are not happy with the selection, you can continue to add points. However, you should know that at some point the results might be less than optimal. In our experience, adding more than 10 even for complex objects often lead to suboptimal performance.

Markers can be moved by left-clicking, holding, and then dragging them to the desired position.

You can remove the marker you placed last by pressing "backspace". You can also select a marker that you want to delete by first clicking on it and then pressing "backspace"

When DEXTR has been activated, you will be able to see a "Threshold" in the top left corner. This threshold can be set from 0-100% by you, the user. It controls how strict the model should be when generating results. The higher the threshold, the smaller and more precise the result will be.

If you are unhappy with your selection and want to start over, you can press “esc”.

When you are satisfied with your selection you can turn it into an object by pressing “enter”, or by clicking the “convert” button in the tool settings toolbar.

DEXTR might take a couple of minutes to activate the first time you annotate as well as when you open up a project where no one has been annotating for the last 30 minutes. If DEXTR is not available to you after 10 minutes, please contact us directly.

Boost model performance quickly with AI-powered labeling and 100% QA.

Learn more
Last modified