Today, for the hackathon we organized at the office, we decided to help blind and low vision people by creating a tool that would help them understand their environment better.
AI in general and image recognition in particular are two technologies that will transform industries as a whole and change many aspects of our daily lives. At its core, computer vision is the ability that machines acquire to see their environment in a way similar as us Humans do. Machines are still far from being able to understand their environment as we do but they can perform very successfully a set of specific tasks in a given environment.
Deepo for good ✌️
At Deepomatic, we use those new technologies on a daily basis to help our clients automate their visual processes or create new products. But today, for the hackathon we organized at the office, we decided to help blind and low vision people by creating a tool that would help them cross the road more safely. At a crosswalk, there is sometimes a tone or a click that indicates when it is safe to cross, but when there isn’t, blind people need to focus on the sound of the trafic coming to determine if they can cross or not.
Helping blind people cross the streets
We are developing a mobile application using the camera of the phone. People will use this application when waiting at a crosswalk, and our technology will tell them in real time the color of the pedestrian light and whether or not there is some trafic approaching.
To build this application, we need to collect as many images as possible. Indeed, that’s a determining factor in the success and the performance of such system. To do so, we decided to call on each and everyone of you and to crowdsource the creation of this dataset, mainly to ensure some diversity in the situations and the locations of the images.
If everyone take one picture, all together, we will have 7 billion picture ???? !
You want to participate? The principle is very simple:
- Take a photo at a crosswalk putting yourself as much as possible in the shoes of a blind person.
- Share this photo on Twitter or Instagram using the hashtag #helpmecross
We will automatically collect your photo, we will anonymise it by blurring faces and licence plates and we will use it to train some deep learning model detecting whether the light is green or red.
Voir cette publication sur Instagram
Thanks for reading, and remember this simple act the next time you are waiting at a crosswalk. You can also share this post to help us spread the word !
If you want to know more about 7 Billion Pictures or Deepomatic, feel free to get in touch with us by email at firstname.lastname@example.org ! Follow us on Twitter ( @7BillionPics ) and Instagram ( @7_billion_pictures) !