DecayDock: The Tiny AI Device That Combats Food Waste

Nobody wants to grab a carton from the fridge and pour a nice, chunky glass of spoiled milk. So when it gets to that point, the only thing left to do is throw it in the trash. It would be much better to eat food before it goes bad, but with a packed fridge, things seem to fall through the cracks from time to time.

Fortunately, there may be a practical way to combat this waste. Instructables user ptallthings93 has created an AI-powered smart fridge companion called DecayDock. It is equipped with a camera that tracks which items are being placed inside the fridge. This information is used to create an inventory system that includes the estimated shelf life of each item. Using this data, the system lets users know what they need to eat before it has a chance to go bad.

The project is designed around an ESP32-CAM module, a compact development board with a built-in camera and Wi-Fi capabilities. Instead of relying on expensive cloud-based AI systems, DecayDock runs its machine learning model locally with some help from Edge Impulse. This allows the device to recognize common food items such as tomatoes, onions, bananas, spinach, milk cartons, and leftovers directly on the hardware itself.

Training the AI model involved collecting dozens of real-world images of food items under various lighting conditions using the ESP32-CAM itself. By avoiding studio-perfect images and instead capturing photos in kitchens and near refrigerators, the system was optimized for real-world use.

When a user places food in front of the camera, the ESP32-CAM captures an image and processes it through the trained AI model. The detected item is then added to a digital inventory displayed on a TFT screen mounted on the front of the device. Freshness is represented visually using color-coded progress bars: green for fresh food, yellow for items that should be eaten soon, and red for food nearing expiration.

To make the device practical for everyday use, the creator designed a custom magnetic enclosure that attaches directly to the refrigerator door. The enclosure houses the ESP32-CAM, display, wiring, and power components in a compact package that resembles a polished consumer product. Powerful neodymium magnets allow it to mount securely without permanent installation.

While DecayDock is still a maker project, it shows how cheap embedded AI hardware can be applied to real challenges in very practical ways. If you’d like to build your own DecayDock, take a look at the step-by-step build guide.