We're still collecting donations
On the 23rd September 2020 we'd raised £220 with 11 supporters in 28 days. But as every pound matters, we're continuing to collect donations from supporters.
Speed Bot presents a way of effortlessly crowd-sourcing traffic sign data, to be made open-source for self-driving car developers worldwide.
by Sida Li in Reading, England, United Kingdom
On the 23rd September 2020 we'd raised £220 with 11 supporters in 28 days. But as every pound matters, we're continuing to collect donations from supporters.
We reached the target! Thank you!
With this extra money, not only will I be able to train the next generation of neural networks for use within Speed Bot, but also implement a remote filtering algorithm for the traffic sign dataset, so that only the highest quality images are made open-source. This will bring us one step closer to introducing novel traffic sign training data to the world! Let's keep this up!
According to the World Health Organisation,
This equates to one death every 25 seconds, and has led to road injury being the 8th highest global cause of death in 2016.
Two significant causes of this are distracted driving and speeding. 46% of cars exceed the speed limit on motorways, 53% on 30mph roads and shockingly 81% on 20mph roads in the UK. Similarly, in the USA, 30% of traffic fatalities are caused by speeding, and this number is dropping by less than 1% every year.
When it comes to distracted driving, 25% of drivers admit to using their phones whilst behind the wheel, and the number of casualties resulting from mobile phone usage is rising year-on-year.
Speed Bot is a free-to-download app on the Google Play Store that aims to reduce the number of traffic fatalities in two ways.
It acts as an adviser on the road. Speed Bot automatically detects, recognises and informs drivers of traffic signs using machine learning, in the hopes of providing a second pair of eyes when necessary. Traffic signs are crucial to a safe journey, as they can immediately provide critical road information and warnings, yet they can be easily missed or ignored in low visibility scenarios.
Speed Bot trumps other AI-based methods in the market, as it is free, efficient (it can reach 8FPS) and accurate. Whilst GPS methods, such as Google Maps, also provide traffic information, sometimes the speed limits shown are wildly incorrect and there is often hours, days or even weeks of delay before they report road works or road closures.
However, Speed Bot's current neural networks are not accurate enough. So a second solution was developed, with a more widespread impact than the first.
As we head into the future, autonomous driving will be more prominent than ever before. Yet with artificial intelligence, there is never a 100% guarantee. AI cannot think like humans do; it relies on the data it has been fed to spot patterns. This means AI is limited by its training data.
Currently, there is a lack of open-source traffic sign data from countries such as the UK, where the traffic regulations are different. For example, we drive on the left side of the road, meaning that the "Keep Left" sign is prevalent. This means that small-scale autonomous vehicle companies without enough globalisation may suffer on the international scale, creating a roadblock to their path to success.
Speed Bot locally stores a selection of the traffic signs detected and recognised. Users through the app can then choose the images labelled correctly, and with a tap of a button, upload these to an online database. Not only will this database have a first-class quantity of images, it will also have an unmatched variety of images from all around the world, as people from 47 different countries have already downloaded the app.
In the future, this database will be made open-source, providing valuable training data for self-driving car manufacturers. Even if this means just one car accident is prevented, there is no such thing as "too safe".
Download the app today, and play a part in shaping the road safety of tomorrow. After all, Little Data can still make a Big Difference.
The money raised through this Crowdfunder will be used to fund the maintenance of the open-source dataset, as well as train bigger and better neural networks. More accurate neural networks means more reliability as an assistant, and more ease as a crowd-sourcing medium. More powerful neural networks with more classes could allow Speed Bot to detect not only traffic signs, but also pedestrians, traffic lights, cars and even more features of the road, providing a diverse open-source dataset perfect for autonomous driving AIs to feed on!
My name is Sida Li and I'm 15 years old. I've always had a passion for maths and computer science, so naturally, I fell in love with artificial intelligence! I initially thought of the concept behind Speed Bot in October 2018, and since April 2019, I've developed the app over 54 versions, and have just released Version 5.00!
Speed Bot was nominated as a finalist of Tech4Good BT Young Pioneer Award. Feel free to share the app and this crowd-funding campaign with your friends and family! Any support is greatly appreciated!
This project offered rewards