A Mexican artist has become involved in an interesting project. Take pictures of your hometown today using Blind Camera, which serves no purpose but creates images through sound. All of this is supported by a neural network running on a Raspberry Pi. To create unique photos impossible with a conventional camera.
One of the most controversial areas facing artificial intelligence is the world of art. Whereas we see developers building learning models to produce artistic results. Mainstream artists are scared, not so scared, of the rise of AI term as they think it could put them out of a job. However, can we consider a camera as an artist? In the case at hand and from a certain point of view, yes.
This is the blind camera that takes photos with AI and a Raspberry Pi
diego trujillo pisanti has created a blind camera that uses AI and is based on a Raspberry Pi that can take pictures. The peculiarity of this model is that it lacks an objective and, therefore, cannot see normally, hence the blind category. Its operation is as simple as the sound produced by the horn, which can be seen in it, using deep learning algorithms based on TensorFlow coupled with a Pi 3 b. All this to produce images like what you can see below these lines.
Despite its low definition, what has been achieved is impressive, especially if we take into account that we are not talking about a high-end PC, but a Raspberry Pi. which have power limits. For creation neural network in charge of images, picenti Used videos from Mexico City to learn the algorithm, Because of this, whatever his blind camera creates, it is reminiscent of that city.
To obtain information with which to train the deep learning algorithm, Pisanty combined the sound produced with the horn of his blind camera and the image of a conventional camera, It could be said that this is a demonstration of how vision works today in living creatures that lack vision or have low vision and are guided by sonar.
Where can I get it?
Unfortunately, due to the fact that the original camera has been used for an artistic project, there is currently only one unit in the hands of its creator and we will not be looking at this as a commercial product. Furthermore, all Deep Learning algorithms are focused on a specific part of the world.
And while the hardware part can be consulted by anyone with a Raspberry Pi and a 3D printer, the entire algorithm for creating the images will need to be re-trained at every location in the world. And you? Do you dare to make your camera blind to see what images you can get or do you consider it a waste of time?
