Object Detection from the Video Taken by Drone

In the last article, we learned how to import the VisDrone dataset into Nilvana™ Vision Studio. It’s not that difficult, just follow the instructions or ask us for help. If you have not read it yet, you definitely have to!

As you can see in the Vision Studio screenshots below, you can clearly see the 10 categories annotated in the images. Some time ago I found a beautiful video — World’s longest drone fpv one shot made by New Step Media on Youtube. I found that there are a lot of indoor scenes that are very suitable for adding the COCO dataset based on the visdrone dataset, so I used this combined dataset to train a object detector. In case you didn’t already know, Nilvana™ Vision Studio has an intuitive web interface that makes it easy to merge annotation data or categories from different datasets with a few mouse clicks.


Visdrone dataset on Nilvana™ Vision Studio
You can watch the finished video below. Kudos to New Step Media for creating such a stunning aerial video. Ownership of the original video belongs to New Step Media.
Drone video object detection