Hornet Detection AI Developed based on YOLOv5

In recent years, there has been a growing concern over the presence of hornets in various areas, posing risks to both people and the environment. To address this issue, I have been working on a solution that utilizes computer vision technology to detect hornets in real-time.

I explored the possibility of building a cost-effective detection system using existing CCTV streams, which can be accessed via RTSP, and leveraging Jetson hardware for processing. This approach allows for efficient deployment without requiring expensive or complex equipment.

To validate the concept, I first created a mock test environment for the project. This involved simulating various conditions to test the hornet detection capabilities under different scenarios. In the test environment, I developed a tracking system by incorporating 9 reference pillars on the ground, mimicking the structure of a real-world environment where hornets might be detected.

The detection model, implemented using YOLOv5, is designed to accurately track and identify hornets in the video feed. The video shown demonstrates the system’s tracking capabilities in action, using the simulated environment. The goal is to further develop this system for real-world applications, ensuring it can provide timely and accurate detection of hornets, helping mitigate potential risks.

It would be fascinating if there were a way to differentiate “killer bees” from other bees and insects just with video. I don’t know anything about it, but I have no doubt the average person cannot identify the variety just by looking at them. If there is some slight difference in appearance between the two varieties, and if you could get images which show those differences for training, then you might have yourself an automated “killer bee tracker”. Imagine the possibilities!