Traditional Industries Embracing AI Technologies

Human evolution since hundreds of years has arrived till this age of Artificial intelligence. Industries have undergone drastic change as well. Just imagine, what would have been a major industry during the era of ancient Indus Valley Civilization? Bricks and pot manufacturing maybe! Industries have kept transforming themselves with a variety of innovative technologies. This was evident in the development of steam engines during the Industrial Revolution in the 1800s or even in the installation of the very first moving assembly line for mass production of an entire automobile by Henry Ford in 1913.

 

In today’s world, industries like software, Internet-based services, travel, communication, electronic items and health care are known to deploy the trending technologies like IoT, deployment of AI in a variety of tasks like prediction, classification and regression among others. But what about traditional industries like agriculture, mining, railways, footwear and food production? Do they have the vision and will to technically evolve along with the rest of the new age industries? The answer is pretty much a loud YES!

At Ignitarium, we have already been working on a whole spectrum of compelling problem statements from both new age and traditional industries. It gives us immense pleasure to share a few of the interesting use cases with respect to traditional industries.

Trains! Right from Dr. Sheldon Cooper from The Big Bang Theory to a little kid from a remote village in India, everyone is fascinated with trains. But are you aware how much of capital and human efforts are spent on maintenance and inspection of railroad ties and plates? That’s where our AI-based solution comes in to solve this complex problem. Our expertise in the domain of image processing plus computer vision has helped us tackle this scenario. We combine aerial imagery with state-of-the-art vision-based AI solutions together to do wonders.

 

As you can see in the image, we are able to detect cracks on the wooden ties. We can also detect missing nuts on the plates. In addition, we have demonstrated our capabilities in detecting other classes of defects such as skewed ties and foreign objects on or near tracks.

One interesting use case is from the humble footwear industry. Nobody wants their footwear less than perfect. Thus, companies have really strict manual quality control personnel whose jobs are to manually inspect cuts, creases and scratches on shoes. We have simplified their jobs with our solution. We have developed a custom neural network to do semantic segmentation that is able to identify the aforementioned defects from a sole in seconds. It can differentiate between a required hole/cut versus a faulty cut. And the good news, the worker gets to keep his job as the software’s GUI is very easy and intuitive for even a semi-skilled worker to operate confidently.

We also tried our defect detection skills to solve a very ‘delicate’ problem. A prototype is built on demand from a reputed customer from eastern part of the globe to crack the problem of vision-based defect detection on eggs! Yes, you read it right. Our system is fully equipped to classify eggs based on color, visible stains and cracks.

 

 

 

 

Similarly, we tried our hand at cookies. In the case of cookies, classes of defects are obviously different. Here the requirement was just to say whether a cookie is perfectly fine or not. A cookie is ‘bad’ if it is broken or over roasted. As you can see, it is a simple neural network with binary image detection capabilities in action.

As you have read so far, we have successfully tackled problems from various industries like transportation, footwear and food . But examples don’t stop here. We are also working in the energy sector. Usage of renewable energy sources like solar power is increasing even in
developing countries. Solar panels are made of many solar cells. There are a number of defects which can be detected visually like micro cracks, scratches, cross cracks etc. Typically, an Electroluminescent camera is used to capture the images of solar panels. Then comes our
custom processing chain which detects the area of the whole panel, separates out individual cells and then classifies the type of the defect.

 

As you can see, cells with very fine scratches are highlighted as defects.

These AI based solutions are getting more intelligent because of trending research in the domain. Vision based solutions have capabilities to surpass human performance. Silicon manufacturers are providing fancy embedded platforms with AI accelerators on board. Cloud
based solutions are also pretty useful at many places where crazy high throughput is not required.

With recent unfortunate pandemic, it is forecasted that there is going to be kind of ‘de-centralisation’ of manufacturing lines from a certain country. Almost all the industries want their manufacturing facilities spread equally across the globe. World will soon come out of gloomy
days. Existing manufacturing lines will run at max performance to complement the lagging supply chains. New lines will be established to tackle the unforeseen hazards.

At this point, comes the opportunity for these traditional industries to try out ‘something new’. There is a well known Chinese saying, “When the winds of change blow, some people build walls and others build windmills.” Let’s build some amazing windmills together!