A picture is worth a thousand words, but is there a hundred words in a hundred words? Creator and developer Diego Trujillo Pisanti sought to find out in his research. raspberry pi-powered blind camera plan. Instead of using lenses to capture light and create images, it listens for sound and builds an approximation of what might be around you based on the detected audio.
This device works like a regular camera in that you point the camera at something you want to capture and press a button to generate an image. However, in this case, it has a giant horn on the front to amplify the sound capture. The user must aim this horn in the direction they want to capture before pressing the button. The camera then analyzes the audio through AI filters to generate images.
According to Pisanty, he developed a custom artificial neural network (ANN) just for this project. He trained his AI using his own model based on a series of videos he shot around Mexico City.
The model was created by taking each frame of video and appending the last second of audio. This helped build sound-video associations that the system could use to create images. Since it was trained in this way, everything it creates is loosely based on the image of downtown Mexico City.
image 1 of Four
In this project, Pisanty uses the Raspberry Pi 3B module. However, it is possible to reproduce using a Raspberry Pi 4. As long as your Pi can handle Tensorflow, it should work. The camera also has a small screen that acts as a kind of viewfinder. This will let the user know that the image is being processed and will show a preview of the generated image. Everything is housed inside a custom 3D printed shell.
The AI model used to train the blind camera was written using Python 3. It is designed to work with Tensorflow 2 and runs on a Raspberry Pi using TFLite. If you want to recreate this Raspberry Pi project, or learn more about how it all works together, check out the video Pisanty shared. YouTube For more information on the project, please visit his site. website.
