Image and Video object detection using YOLOv3 (https://pjreddie.com/darknet/yolo/)
I think you must have been living under a rock to not know about the Lawyer on a Zoom call that couldn’t turn off his Cat filter and declared…
This repo Leverages YOLOv3 image detection to classify objects in images and videos. “You only look once (YOLO) is a state-of-the-art, real-time object detection system.”
https://pjreddie.com/darknet/yolo/
Photo object detection is quite effective. To execute the python script using system arguments, use the following syntax via CLI:
python yolo.py —image images/beagles_roof.jpg —yolo yolo-coco
Below is an initial image object detection using a photo of my dogs. The classifications are pretty spot on (no pun - by the way, their names are Moe and Bodie).
Let’s take a look at another image, this one with more quite a few more obvious objects.
The algorithm is 99% certain about the dining table but could not assign a classification to the pair of glasses, the pen, or the Bubly can (these three classes aren’t included in the pre-trained COCO set). Even the bowl way in the background was detected. Lowering the confidence parameter down to 10% will pick up the can as a “cup”.
The video object framework works quite well in real-time on videos. For testing, a collage of airport footage was fed through. Below is the classification output:
The icing on the cake is extracting data from the frames for analysis. Below we can see number of frames for each object that was detected as well as confidence.
