SMART PET DOOR

My lab partner and I were tasked with designing a motorized "doggy door" that opens only when a recognized pet wants to enter a home. If, for example, a stray dog attempts to enter, the door should remain closed. Project stipulations also required that this be done using only a Raspberry Pi and 3D-printed parts.

Though machine-learning has become much more accessible in recent years, it still often requires powerful hardware to have any viable applications in industry. In addition, the steep learning curve associated with training a model makes the power of ML difficult to harness as a newcomer. 

The current version of the pet door is both quick and reliable. All that is required are a few images of the user's pet. Using these images, a TFLite model is generated and run using just a Raspberry Pi and Coral USB accelerator.