Colour By Column in Zegami

We’ve added a simple and powerful new feature to Zegami: Colour by Column. 

Colour by Column enables each item in the Zegami viewer to be tinted in a colour, which reflects the value of any chosen column. This could be a score, a date or some other metric, or a category/label. You can simply pick a column, and every item in the view will immediately take on a tint to show its value for that column. 

This adds a powerful new dimension to visual data exploration in Zegami, enabling richer views, which reveal more of the underlying patterns in your data. Because the image itself also shows through, the full benefits of Zegami’s engine are retained, enabling you to zoom right in and see the full detail of each picture. 

Colour By Column works equally well for numeric values or if your values represent categories. 

Here are a few examples: 

Heart Disease 

In the below screenshot, we have run an unsupervised machine learning model to cluster images. By overlaying Colour by Column, showing whether the heart was part of the control group (blue) or diseased (orange), we can immediately see how the scans with disease form a distinct group. This helps doctors identify disease when new heart scans show similar characteristics. 

Unsupervised machine learning from echocardiograms

Red Bull Social Media 

As another example, we can look at our Red Bull marketing images collection built for their Instagram account. Once again, we have run some machine learning models to cluster the images based on similarity. 

This time we will use Colour by Column to tint the images by date. As the date is a continuous variable, we can tint the images from oldest to newest, red to blue respectively. We use a stronger colour effect here because the images themselves vary significantly in their own colour. 

Similarity clustering of Red Bull’s Instagram images

We can see that posts with earlier dates were more tightly clustered around one part of the similarity space. This indicates that the original manager of Red Bull’s Instagram tended to post visually similar content, whereas later posts have a much greater variety. 

This view can then be pivoted to show how the likes have been affected by this strategy shift and immediately we can see that the newer posts get many more likes indicating the strategy is working well.