This is an implementation of an object detection system built on RetinaNet, using Tensorflow 2.0, to detect birds and rats. Possible use case in outdoor food establishments.
At present in some local coffeeshops / food courts in Singapore, some inconsiderate customers may have yet to catch onto the whole 'return your own tray' movement, resulting in uncleared trays left unattended.
Often when this happens and there are insufficient cleaners to clear the food trays, our local avian friends (AKA Javan Mynah and Common Pigeon) would swoop in for a quick easy feast on the scraps of leftover food. While seemingly philantrophic, this has severe repercussions on hygiene and food safety standards as these birds may pass on bacteria and viral infections.
In addition, rat infestation is a taboo in local food centres as rats are one other carrier of bacterium. Often, sightings of rats in food establishments would lead to its temporary closure, for thorough washing and disinfection.
Thus, I propose installing a robotic laser pointer! Folks working in these food courts have been known to be able to scare off these birds by pointing their laser pointers at them. With a robotic laser pointer armed with a simple object detection system (Pest Detector), the task of scaring off these birds could be easily automated! On the other hand, if rats were to be spotted, a video could be logged and flagged out as a warning.
Model used: RetinaNet Github repo here
Dataset source: Open Images Dataset v6
Test video source: Video 1 | Video 2
Done by: Cheo Kee Jin