Introducing a revamped Picterra UI
Back when we first started developing the Picterra Platform, our UI was based purely around the idea of annotating images. This was even before we had the custom detector. Admittedly, it had some rough edges and some hard to understand interactions. Since then however it has been evolving and improving, allowing users to easily and intuitively harness the power of the platform for themselves.
In the last few months, with the help of our new UX designer, we have been rethinking our UI and finding ways to put detectors at the center of it. We also designed it in a way that we can simply extend this same UI as we allow our users to scale their detection to very large images or WMS. In order to do this…
We made the separation between training and running a Custom Detector clearer. The new UI now has a training mode and a detection mode.
Today, we are rolling out a revamped UI for all of our users !
The Dashboard
The focus is now on detectors, with the landing page of the platform being your Dashboard. From the Dashboard, you can access your detectors or your imagery:
For existing users, the main changes is that the notion of ‘project’ is gone and it has been replaced by folders in your imagery. All existing projects have been turned into folders.
Training mode
When managing your detectors and results, your interaction will be centered around different modes: Training mode, Detection mode and Stats & Report mode. The mode is visible in the black bar at the top.
In the previous UI, Custom Detectors also belonged to a project, which means you could only use it from within this project. This is not the case anymore and you can now use and train your Custom Detector on any image from your library.
When you create a Custom Detector, you will see a screen where you need to select which images to add to your detector. Those images will then be used for training your detector.
From the Training detector mode, you can add additional images by clicking on ‘Add my images’ in the Images widget on the top right of the screen.
Once trained, you can either improve your detector or run it on other images, in order to apply it at scale.
If you need more tips & tricks to train a good detector, check out this article.
Detection mode
After you spent some time training your detector and you have reached a good accuracy level, you can to run it at scale. The new Detection mode allows you to run a detector on up to two images in parallel (this number will increase in the near future).
In addition, it allows you to run the detectors in the background, so you can train another detector while your first one is running on your imagery.
Stats & Report mode
After a detector has finished running on an image, you get an indicator on the ‘Results’ button next to that image:
The Stats & Report mode allows to view the results across all detectors for a given image, view some stats about them, generate a PDF report and download the results in your format of choice (CSV, GeoJSON, Shapefile).
What’s next ?
Besides the changes that we introduce today, the other goal of this new UI is to lay out strong foundations for what is to come next. In particular, we want to focus on two things:
- Running detectors at scale, either on large WMS regions or on very large GeoTiffs
- Bringing new tools to evaluate the accuracy of your detectors
We would also like to get your feedback on this new UI, so please do not hesitate to reach out by email or in the comments.
Start testing it by registering on picterra.ch !
Last but not least, kudos to our UX & engineering team for the hard work in the past few months !
