This DIY camera uses machine learning to tell you what it sees

A camera that knows what it’s looking at would have seemed like a farfetched pipe dream not long ago. These days, however, you can even build such a camera in the comfort of your own home. Adafruit has shared a project that allows you to build a camera around Adafruit’s own BrainCraft HAT system, which is itself an AI Machine Learning addition to the Raspberry Pi 4. Plus, it even tells you what it sees.

As DIY Photography observes, the Raspberry Pi Machine Learning camera uses TensorFlow Lite object recognition software to figure out what it is looking at. The project works with either the 8MP Pi Camera Module or the 12.3MP module that includes interchangeable lens support. The enclosure is 3D printed.

So what is a camera that can determine what it’s seeing be used for? As John Aldred points out at DIY Photography, ‘…it opens up a lot of options for connecting the Pi to a ‘real camera’ for shooting photos or video. You could connect a DSLR or mirrorless camera from its trigger port into the Pi’s GPIO pins, or even use a USB connection with something like gPhoto, to have it shoot a photo or start recording a video when it detects a specific thing enter the frame.’

Imagine a scenario in which you want to remotely photograph a specific animal. Perhaps you don’t want to have a camera constantly capturing images or shooting whenever it detects motion, but instead want something more precise and fine-tuned. Maybe a sports photograph could utilize object recognition to capture a specific moment during a game with one camera while manually shooting with another. Aldred also notes that you could potentially have the object recognition software place object keywords into image metadata, which could make organizing large batches of images a much more streamlined process.

Image credit: Adafruit

The technology is still developing, and machine learning takes time, but there’s a lot of interesting potential here and talented, creative individuals often find ways to push new gear and tech to the limits. At the very least, it seems like a fun project. If you’d like to learn more and view the complete parts list, click here. In total, you’ll need the Adafruit BrainCraft HAT (around $40), Raspberry Pi 4 ($35), a camera ($30 and up) a 3D printed case, and a few more odds and ends. Click the following links for helpful user guides: ‘Easy Machine Learning for Raspberry Pi‘ and ‘Running TensorFlow Lite on Raspberry Pi 4‘.