The humble Raspberry Pi single board computer (SBC) has been featured in many breakthrough manufacturing devices. Rightly so, and in the decade since the Raspberry Pi was announced, the community of makers has grown steadily. Low-power SBCs like the Raspberry Pi provide the ports and intelligence. We don’t often see the light shining on DIY maker failures, but as the 8-bit, one-byte video below (surfaced by Hackaday) demonstrates, they can be just as worthy of attention. The 8-bit, one-byte AI tag maker meets a real need for well-organized people, is quirky and fun, and allows us to learn how to create useful devices to move from theory to reality.
Labeling is usually a time-consuming process, but a particularly interesting activity in a tech and DIY fan’s house, with all of their tools, components, parts and parts, tidy and filed all around the house, from the hangar and / or garage. It is also an activity that can help with the day-to-day organization of the home, which, according to 8 bits and one byte, “is the new trend”. Perhaps it is Marie Kondo’s tidy influence seeping into tech fans.
You can view the full list of hardware supplies and the code needed to get this Ai labeling project up and running through its associated Unstuctables page. In short, the project as seen in the video includes a Raspberry Pi 3B +, a Raspberry Pi camera, an Adafruit Mini thermal receipt printer and an arcade button to activate the camera, and some cardboard origami for housing the components and making one to handle the device. The key API running on the device is the DeepAI DenseCap API. This API analyzes images in the DeepAI cloud and returns the recognition result as text for thermal printer output.
AI meets Murphy’s Law and complex image backgrounds
The first job of any self-respecting AI label maker is obviously to label himself. 8 bits and one byte found a “mirror tile array” for this purpose, which did not achieve the best result, printing a “window is white” label. Not a good start, but the tiled mirror was not optimal.
A portion of the video titled “Labeling All Things” didn’t really prove the merit of this AI tag maker, either. An apple was then labeled “the man’s head”, a mug “the windows are open” and chocolate cookies were “a cell phone”.
The success rate of AI obviously requires some attention. We expect observers to quickly conclude that there is a neutral, pristine backdrop behind 8-bit objects and a desired byte tagging would have helped the AI a lot. It can be a little frustrating but gives some hope to label tinkerers, believing that better results will be possible, based on the work presented in the video and in the Instructables documentation.