Object Identification with Raspberry Pi

Necessary materials:

When we press the button, the camera takes a photo. The captured image is sent to TensorFlow and the image is processed. As we mentioned before, TensorFlow tries to make sense of the image and makes 5 guesses about what the image might be. Generally, the top 2 estimates are the closest. We will translate the first of these predictions from text to audio with Flite, which is a text-to-speech application. We will print all of the predictions or as many as desired on the LCD Screen. We printed two.

Since there is no device that can output sound on Raspberry Pi, we need a speaker. We also used the DIY Speaker Kit to hone our soldering skills. You can buy this kit if you want, but any loudspeaker will do. You can even wear headphones.

Our LCD Screen also comes with separate headers. After soldering the Header to the pins of the LCD Screen, we solder a 1k Ohm resistor to the 3rd pin of the LCD Screen as in our Digital Meter article.

Then we plug the Speaker’s USB cable into one of the Raspberry Pi’s USB inputs. Our speaker will get its power from here. We plug the 3.5mm jack of our speaker into the 3.5mm jack input of the Raspberry Pi.

Let’s connect our camera to the Raspberry Pi with the blue band of the ribbon cable facing the USB ports as in the photo.

Friends who want to put their projects in the box as we did, can make the connections as follows to save space. (NOTE: Although the LCD screen sizes are different, the connections are the same. However, some changes are required in the code part to use screens of different sizes. We used a 20×4 LCD screen. )

Leave a Comment