Keeping track of images and their corresponding annotations is a challenge. Knowing which annotations map to which class, viewing image metadata, and seeing which images correspond to a training, validation, and testing set can quickly become a sprawling mess.

We're introducing improved detailed image previews that enable our users to more easily see individual annotations, detailed image information, and more.

Let's walk through an example from an aerial object detection dataset.

Images and their annotations are now displayed full screen. Annotations can easily be toggled on and off by mousing over top of the eye icon.

Use Roboflow to view images with or without annotations.
Toggle annotations on/off by simply mousing over top of the eye icon.

Moreover, single objects can be viewed by simply mousing over individual annotation groups. Here, we view the lift, jetski, dock, and boat independently.

Use Roboflow to view individual annotations.
Mouseover an annotation to see it individually.

Even better, Roboflow easily surfaces the image metadata and details, so you can see attributes like the original image name, when it was last modified, the camera used to capture it, and if that image is in the training, validation or testing set.

Use Roboflow to understand image metadata and details.
View image details and metadata easier.

If that image is in a version that you applied augmentations, the details also contain the individual augmentations applied to that given image as well. For example, if you'd like to see the amount of brightness added, that is available.

We've heard from our users that these changes make it much simpler to navigate their image dataset. As always, we welcome feedback on how this can be even better:

Happy building!