As you can see the previous blog post, we will add more and more features to meet the needs. Today, we would like to show a sample of computer vision outputs that you should expect soon. The feature is still under development and still being tested.
The below photo is from Nepal earthquake. As you can see, the features are description, tags, category, adult content and racy content with scores. It gives us information that this image is related to disaster or not. In addition, it gives some aspects of an image. See tags section, it contains bed, blank, colourful. It indicates a shelter.
If the computer vision output gives us a muddled information with low score, we will push these data to our clickers. Then, input from clickers will teach our computer vision algorithm to generate better output with higher score.
The service will be available in the early of March on selected data collection only. Next blog, we will show how we are resolving damage assessment on aerial imagery using deep machine learning.